Cappels' project pages
Return to HOME
Making Matched Pairs And Sets Of Transistors
A fixture and a method of matching the VBE of
The hardware is pretty simple: A power
supply, a few resistors and some sockets.
The magic is in the
technique, and its all pretty simple and straight forward.
schematic shows two pairs of sockets, one for NPN and the other for
Realizing that I will also need mathced PNP transistors for a
I added the PNP sockets after the photographs on this
page were taken.
Find updates at www.projects.cappels.org
I needed an AC wattmeter, and while collecting bits an pieces for
something based on a commercial wattmeter chip, I came across and
elegant design in a National Semiconductor application note,
Application Note 222, "Super Matched Bipolar Transistor Pair Sets
New Standards for Drift and Noise", July 1979, featuring the
LM194 dual transsitor. That was a long
and unfortunately, these parts have been out of production for a
long time. According to their website, Analog Devices still makes
a matched pair of transistors, but I was unable to locate a source of
them at a reasonable price, and this is important because I buy lots of
spares because I tend to destroy a lot of parts, especially in power
circuits, and therefore decided to make my own matched
wattmeter project is described on this web page.
Bob Peas, at National Semiconductor expanded on the design of the
original wattmeter, and recognizing the problem of finding matched
pairs, wrote a short article, How
Your Own Matched Transistors, which unfortunatley is not
available on the internet at the moment, possibly lost when the
National Semiconductor website was moved to Texas Instruments. The
articles referenced within Mr.
Peas' article have also disappeared from the web, but I think I found
them and have
duplicated the information on this web page. Most of this article is a
rehash of the information provided by National Semiconductor with
original of my own added other than some pictures and an account of my
The transistors I decided to use for this is Fairchild's 2N3904. This
also dates back to the 1970's, but its still being made and is very
inexpensive. I bought 100 2N3904's for $US2.47. From this 100 parts, I
screened approximately 80 of them, and from those 80, I was able
to match the VBE of more than 60 of them to another transistor to
100 microvolts. If I were to throw away the unmateched
transistors, each matched pair would cost me only 8.2 cents (US) each,
not counting the labor.
After matching a batch of 2N3904's, I decided to make some matched set
of 2N2906's, so I added a second set of sockets to accomodate them.
This simple fixture can also be used to compare and match the forward
drop of diodes at 1 milliamp. Of course if you want to test at a
different current, change the resistors but remember to take the diode
drop into account.
The only purpose
for the LED is to let me know when that the power is on.
The circuit is about as simple as it can get. +8 to +24 volts is
applied to the power input connector. A 78M05 regulator reduced and
regulates the voltage to + 5 volts. A lower power regulator could be
used, but I have many 78M05 regulators and size was of no particular
concern. The transistors under test are diode-connected (that is to say
that their bases are connected to their collectors). In the case of the
NPN transistors, the emitters
grounded, the collectors and bases of each transistor are connected to
+5 volts through a 4.32k 1% resistor. The Base-Emitter voltages of the
2N3904 transistors that I was sorting was approximately 0.68 volts at 1
milliamps, so the current through the transistors is (5.0 volts - 0.68
volts) / 4.320 k Ohms = 1 milliamp exactly.
Why didn't I use a constant current source? The voltages of the
junctions are all similar and the slight mismatch would result in a
very small difference in total voltage across the resitor, so the 1%
tolerance of the resistors will dominate the current difference. Since
the voltage drop across the transistors is a function of the log of the
current, the difference in the voltage resulting from the mismatched
current is insignificant when matching to within 100 microvolts.
The 7805 that I used provides 4.98 volts, making the current pretty
accurate. If you have a strong desire for more accuracy, you can select
a regulator that is closer to 5.00 volts or use an adjustable regulator
such as an TL431 or LM317. Accuracy of the test current is not very
important. What is important is that the currents though and the
temperatures of the two transistors under test be as close to being the
same as possible, so matching the 1% resistors might be a better
investment of time.
Rather than measuring the voltage across each transistor, which would
require one more digit on my voltmeter to obtain 100 microvolt
resolution, one reference
transistors is arbitrarily selected and put into the "Reference" socket
and then the transistor being sorted is put into the TUT (Transistor
Under Test) socket. The voltmeter is connected between the two
transistors with the negative lead connected to the base and collector
of the reference transistor and the positive lead connected to the base
and collector of the TUT. Thus, if the meter reads 1.2 millivolts, it
means that the base-emitter voltage of the TUT is 1.2 millivolts higher
than that of the transistor under test.
The ground test point is there to allow measurement of the actual
base-emitter voltage. Sometimes that is interesting or helpful to know
that, such as when diode-connected transistors are used as temperature
probes, direct measurement of that voltage is necessary.
I did not bother with any fancy packaging -not even a box. After all,
transistors need to be in the open for access to the sockets and so the
fan can blow on them.
The sockets I used are some that I found in a surplus store with female
contacts 2.54 mm (100 mill for Americans) centers. They have
leaf contacts and though I don't know what for what use they were
originally intended, I use them with Berg Connectors. You can also use
transistor sockets. I cut down the connector I used so there were two
rows of three pins. The spacing for Berg Connectors is idea for
mounting two TO-92 transistors with their flat faces facing
one-another. I think that when I use matched pairs in a circuit, they
will be mounted this way; with the flat faces toward each other, most
likely coupled with high therma conductivity epoxy, so arranging the
sockets in this orientation will make it easier to test pairs of
transistors that are glued together.
The setup as on the seat of a chair,
which makes a clear and convenient
workspace in the midst of workbenches overflowing with junk.
To sort the transistors, set up on table reasonably free of clutter to
allow for free airflow over the surface.
• Place a fan so that it constantly blows room temperature air across
the Reference transistor and the Transistor Under Test. This helps both
transistors settle to the same temperature quickly.
• Attach the test leads to the + and - terminals on the fixture so that
you will not have to hold the leads in place.
• Keep the transistors that are about to be tested in the same room
temperature stream of air from the fan so that they are kept at close
to the same temperature as the Reference Transistor, making settling
• Use a pair to tweezers, and your warm fingers as little as possible,
pick up the Transistor Under Test and place it in the test position.
Keep the tweezers in the fan's output stream to help keep it at the
same temperature as the Reference Transistor.
• Wait as long as you can stand to, to make sure that the voltage
across the Transistor Under Test has stabilized. This will be indicated
by the voltage on the voltmeter settling down.
I the case of my first use of this fixture, the air temperature in the
room was 24.2° C, Whenever I picked up a transistor with my
fingers, if even for a one or two seconds, the temperature of the
transistor would increase by two or three degrees. It would take a
painfully long time for the transistor to settle down after touching
it. Keeping the tweezers laying in front of the fan instead of in my
hand and keeping the transistors spread out in front of the fan, rather
than in a pile as shown in the photograph, often reduce the settling
time to a few seconds.
If you see this webpage on a site other than cappels.org,
please email me at the address below. Thankyou.
This picture was taken after I had already bagged and labeled ten
matched to within 100 microvolts ofthe Reference Transistor - 0.1 mv
I decided to bin the transistors at 100 microvolt intervals, which was
the resolution limit of my digital voltmeter, even though my initial
application can tolerate up to approximately 400 microvolts of
difference because there were enough matches to have paris more closely
matched, and sometime in the future, this closer matching might be
useful. After binning them, I further sorted two of the bins into
"upper part" and "lower part", by picking one transistor from each bin
and comparing the other transistors from that bin. That allowed me to
be assured that those that went through secondary screening are within
100 microvolts of one-another.
It was both surprising and pleasing to see that in this batch of
Fairchild 2N3904 transistors, the all matched one-another within two or
three millivolts at the most, and by far the bulk of them were within 1
millivolts of the arbitrarily chosen Reference Transistor.
Matched Transistors In Pairs
Back in the late 1970's I designed video camera deflection circuits
that used a pair of 2N3904 transistors as a wide bandwidth differential
amplifier. At that time, I thought the best method of thermally
coupling the transistors was to apply Wakefield thermal coupling
compound to the faces of the transistors and then fasten them with a
Panduit Tie Wrap. Today, having the transistors face each
other is still a good idea and a plastic wire turns out to be a very
To couple two transistors together, I put two transistors from one
±100 uV bin into the base-emitter voltage test fixture and let
the temperatures of the transistors stabilise to verify that the
transistors are indeed matched. This precaution is taken because
sometimes mistakes are made or an error in the original measurement
could have been made, and it seemed prudent to verify the quality of
the match before investing a lot more time and resources in using the
Once it is apparent that the transistors are matched, keeping the
transistors in the matching fixture, add a drop a fingernail varnish to
the face of one transistors, and then fasten the two transistors
together with a plastic wire tie. Put more fingernail varnish over the
top of the transistors and to where the wire tie touches the
transistors to keep the whole thing from slipping apart. I used green
fingernail varnish to remind me that these are NPN transistors. I used
red for the PNP transistors. Allow the varnish to dry and the
temperatures to stabilise once again, and check again to verify that
the base-emitter voltage drops of the two transistors are still
A useful side effect of using a plastic wire tie is that it offers a
little bit of thermal insulation so as to reduce thermal gradients
generated by air currents. In use, copper foil was wrapped around the
pair to give the device more themal mass so that the temperature would
not change as quickly, the pairs are placed inside a
small circuit enclosure or embedded in a block of plastic foam to
reduce the effects of air currents on the offset. If the leads are kept
short, the leads won't be affected by air currents and the circuit
board can act as additional thermal coupling between the two junctions.
Note: Thanks to Brian in Pittsburg, PA for letting me know about an
error in the text.
Contents ©2011 Richard Cappels All Rights Reserved. Find updates
First posted in January,
2011 Updated December 28, 2011.
You can send
email to me at
"(at)" with "@" before mailing.
presented on this page is for personal, nonprofit educational and
use only. This material (including object files) is copyrighted by
Cappels and may not be republished or used directly for commercial
For commercial license, click
and intellectual property notice
(Summary: No warranties, use these pages at
own risk. You may use the information provided here for personal and
educational purposes but you may not republish or use this information
for any commercial purpose without explicit permission.) I neither
express nor imply any warranty for the quality, fitness for any
particular purpose or user, or freedom from patents or other
restrictions on the rights of use of any
software, firmware, hardware, design, service,information, or advice
mentioned,or made reference to in these pages. By utilizing or relying
on software, firmware, hardware, design, service,information, or advice
provided, mentioned, or made reference to in these pages, the user
takes responsibility to assume all risk and associated with said
activity and hold Richard Cappels harmless in the event of any loss or
expense associated with said activity. The contents of this web site,
unless otherwise noted, is copyrighted by Richard
Cappels. Use of information presented on this site for personal,
educational and noncommercial use is encouraged, but unless explicitly
with respect to particular material, the material itself may not be
or used directly for commercial purposes. For the purposes of this
copying binary data resulting from program files, including assembly
code and object (hex) files into semiconductor memories for personal,
educational or other noncommercial use is not considered republishing.
desiring to use any material published in this pages for commercial
should contact the respective copyright holder(s).