Transistor matching--- Vbe or Hfe
r.l...@comcast.net
I normally match pairs using the Hfe method. But this would be a repair
situation only.
But soon I'll need to do matching of some TO126's on a small production
scale.
I also will be matching some TO92 current mirrors.It seems to me that Vbe
would
be the more important requirement. I can't see matching both Vbe and Hfe.
I don't think you'll find many pairs that match both parameters. It would
also be
very time consuming. What do you guys recommend. I'll be building a test
fixture for the test
and can build what I need for the project.
Any input appreciated.
RonL
John Popelish
It all depends on what will be driving the matched pair-- a
common voltage or two equal (or a common) currents. For the
current mirror, it is almost certainly a common voltage.
That goes for differential pairs, also.
--
Regards,
John Popelish
Phil Allison
"John Popelish"
>
> It all depends on what will be driving the matched pair-- a common voltage
> or two equal (or a common) currents. For the current mirror, it is almost
> certainly a common voltage.
> That goes for differential pairs, also.
>
** Huh ??
A diff pair of BJTs needs to have matched gain ( Hfe) transistors as well as
( ideally ) matched Vbe.
With unmatched Hfe devices, input offset current can become very
significant resulting in large DC offsets at the output of the circuit that
could only be fixed with an offset trim control.
...... Phil
mike
Calculate the cost of the fixture, training, redesign, repair...
+ the labor cost of the matching
+ the inventory/distribution/repair cost of managing another part
number of pairs
+ the cost of redesigning the circuit when you get tired of all the
matching issues.
Divide by the quantity.
Compare that to the incremental cost of buying a matched dual.
If your time is expensive and your production run is "small"
you've probably already wasted more $$ than it would have cost to
do it right the first time.
Horror story.
One company saved 20-cents by using matched fets and poor layout.
Somebody decided to improve system cooling.
The resultant changes in heat distribution shut down
the production line 'cuz the fets didn't track any more.
Sounds like you've got a temperature-sensitive situation.
Recipe for disaster.
It's easy to see the 20-cents extra cost. What's hard is the
cost of fixing all the problems it causes.
--
Return address is VALID!
Bunch-O-Stuff Forsale Here:
http://mike.liveline.de/sale.html
Robert Baer
If you are making current mirrors, Vbe is the only important parameter.
In fact, one can mismatch Vbe by a specific value to give a 2:1 or
3:2 or some other ratio.
In this manner, emitter "balancing" resistors are not needed, and the
current ratio is kept over a decently wide temperature range.
Jim Thompson
Amazing! What is this "decently wide temperature range" you speak of
?:-)
...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: "skypeanalog" | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
America: Land of the Free, Because of the Brave
Robert Baer
Jim Thompson
<rober...@localnet.com> wrote:
The ONLY way you can do that is if you can control IS. Picking
transistors out of the bin, Vbe mismatch is likely due to uneven
emitter depth and/or metalization flaws... not something I'd like to
hang MY hat on.
In the integrated circuit world we create ratio'd current mirrors by
sizing the emitter area, something you can't do with discretes.
Even then, I still use ballasting resistors to accommodate slight
mismatches.
Tim Williams
seem to match well. I've got a couple hundred Fairchild 2N3904's, all with
the same date code, which match within milivolts. I build a pair of
differential amplifiers with a few and, taking no special precautions for
balancing (e.g. no current mirror collector load), I got under 20mV at the
output (input grounded), which is under 2mV at the bases. If I'm not
mistaken, that's op-amp quality balance!
Tim
--
Deep Fryer: A very philosophical monk.
Website @ http://webpages.charter.net/dawill/tmoranwms
"Jim Thompson" <To-Email-Use-Th...@My-Web-Site.com> wrote in
message news:dnam14hc105hla0hd...@4ax.com...
Jim Thompson
<tmor...@charter.net> wrote:
>Transistors from the same batch (and especially the same wafer, I assume)
>seem to match well. I've got a couple hundred Fairchild 2N3904's, all with
>the same date code, which match within milivolts. I build a pair of
>differential amplifiers with a few and, taking no special precautions for
>balancing (e.g. no current mirror collector load), I got under 20mV at the
>output (input grounded), which is under 2mV at the bases. If I'm not
>mistaken, that's op-amp quality balance!
>
>Tim
I think I misread "Deep Fryer's" assertions as selecting a PAIR of
transistors to do 2:1, etc.
If you ratio _two_transistors_ to _one_transistor_ you can, indeed,
make a reliable 2:1 mirror.
My apologies.
Tim Williams
> I think I misread "Deep Fryer's" assertions
- Me? I was just dropping in some experience.
What was the topic, something about current ratios? That was like four
posts ago, who remembers...
Robert Baer
2:3, etc and they all consistently worked.
Yes, i know that emitter areas on ICs are used to get that mismatching.
I better, because i worked in the analog divisionfor Fairchild for 5
years.
The discretes were used in making a stick prototype for datasheet
testing; usually the differential "front end" (with current source) was
"stolen" by use of a metal mask from some other part with the same
proposed diffusion schedule (say the 709).
Some other sections were done this same way.
Jim Thompson
<rober...@localnet.com> wrote:
>Jim Thompson wrote:
[snip]
>>
>> In the integrated circuit world we create ratio'd current mirrors by
>> sizing the emitter area, something you can't do with discretes.
>>
>> Even then, I still use ballasting resistors to accommodate slight
>> mismatches.
>>
>> ...Jim Thompson
> Well, i used to do a lot of discrete mismatching to get 1:1. 1:2,
>2:3, etc and they all consistently worked.
> Yes, i know that emitter areas on ICs are used to get that mismatching.
> I better, because i worked in the analog divisionfor Fairchild for 5
>years.
> The discretes were used in making a stick prototype for datasheet
>testing; usually the differential "front end" (with current source) was
>"stolen" by use of a metal mask from some other part with the same
>proposed diffusion schedule (say the 709).
> Some other sections were done this same way.
Yep. At Motorola we first used metal patterns to create "kit parts"
from existing chips. Later we actually ran kit parts as a mask unto
itself.
Robert Baer
> On Sat, 03 May 2008 02:42:33 -0700, Robert Baer
> <rober...@localnet.com> wrote:
>
>
>>Jim Thompson wrote:
>
> [snip]
>
>>>In the integrated circuit world we create ratio'd current mirrors by
>>>sizing the emitter area, something you can't do with discretes.
>>>
>>>Even then, I still use ballasting resistors to accommodate slight
>>>mismatches.
>>>
>>> ...Jim Thompson
>>
>> Well, i used to do a lot of discrete mismatching to get 1:1. 1:2,
>>2:3, etc and they all consistently worked.
>> Yes, i know that emitter areas on ICs are used to get that mismatching.
>> I better, because i worked in the analog divisionfor Fairchild for 5
>>years.
>> The discretes were used in making a stick prototype for datasheet
>>testing; usually the differential "front end" (with current source) was
>>"stolen" by use of a metal mask from some other part with the same
>>proposed diffusion schedule (say the 709).
>> Some other sections were done this same way.
>
>
> Yep. At Motorola we first used metal patterns to create "kit parts"
> from existing chips. Later we actually ran kit parts as a mask unto
> itself.
>
> ...Jim Thompson
Fairchild.
**
OT (true story):
At one point working there, a rumor was being circulated that
Fairchild and Honeywell were going to merge, and that the new company
was going to be named "Fair Well Honey Chile".
Jim Thompson
<rober...@localnet.com> wrote:
Or the one my 90+ year-old grandmother used to tell: Stop & Shop and
A & P are going to merge and be called Stop & P ;-)
Robert Baer
JosephKK
About half of the semiconductor manufacturers in the 1970's did that.
You could buy many kinds of uncommitted and semi-uncommitted
transistors arrays. In a couple of TO-100 packages or dips (with some
discrete parts) you could a full modern (at the time) operational
amplifier.