Transistor matching--- Vbe or Hfe

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

"John Popelish"

** 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

Do the math. 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

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.

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

Will MIL spec do?

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.

...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

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

-- Deep Fryer: A very philosophical monk. Website @

parameter.

the

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.

...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

- Me? I was just dropping in some experience.

What was the topic, something about current ratios? That was like four posts ago, who remembers...

Tim

-- Deep Fryer: A very philosophical monk. Website @

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

--
| 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

..so, the term "kit parts" was industry wide? That was the term used in 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".

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 ;-)

...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

*MUCH* better!

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.