Why the color bands on TO-92's back side?

Thread Starter

Willen

Joined Nov 13, 2015
333
I have some old TO-92 transistors. Some have color bands/bars behind its face/marking.

For example I have 2N4401 manufactured around 1988 and I am writing their markings and different color bands here (All have N's logo, it means all these were manufactured by National semiconductor):
- Marking: "428 2N4401" & color bands: "blue-red-yellow" (bottom to top) I don't know where to start see bands- top to button or button to top.
- Marking: "428 2N4401" & color bands: "violet-red-yellow" (bottom to top)
- Marking: "416 2N4401" & color bands: "yellow-brown-yellow" (bottom to top)
These bands are mysterious to me. Can any one decode it? (I see no such bands in new parts)
 
Last edited:

bertus

Joined Apr 5, 2008
22,270
Hello,

The transistors seem to be tested for certain parameters and put in to classes.
One of the parameters could be the Hfe.

Bertus
 

Thread Starter

Willen

Joined Nov 13, 2015
333
But I have a pack of 250 pieces of these and approx 20% of them have such bands. So it's hard to believe that they test their 20% (huge amount) of their product.
 
Last edited:

Thread Starter

Willen

Joined Nov 13, 2015
333
I tested their hFE now just for simple guessing. I collected 14 pieces of same color-band transistors. Their hFE variation is large! Some have 160 and some have 360!

Other (like 2N3904, BC547B etc) value of transistors have very near variation on their hFE. 10 pieces of 2N3904 have their hFE between 190 to 210. But I also tested new 2N4401 (by Fairchild) and these also have large variation of hFE. Maybe its chip is not improved till now from few decades. :)
 

Kermit2

Joined Feb 5, 2010
4,162
As an aid to a person assembling the board and as an indication of correct placement for error checking/quality

The sorting by parameters is also done, and what you have may be the rejects from a factory sort bin order vs. what the assembly dept find when they test for the same parameters as the build is in progress.
 

GopherT

Joined Nov 23, 2012
8,009
I tested their hFE now just for simple guessing. I collected 14 pieces of same color-band transistors. Their hFE variation is large! Some have 160 and some have 360!

Other (like 2N3904, BC547B etc) value of transistors have very near variation on their hFE. 10 pieces of 2N3904 have their hFE between 190 to 210. But I also tested new 2N4401 (by Fairchild) and these also have large variation of hFE. Maybe its chip is not improved till now from few decades. :)
Thanks for asking this question, I have often thought about it as well. I have a bin full of transistors from the 1960s/70s according to some notes on the bin - there are no other markings on these old GE and zenith transistors except for the color bars. I assumed that early transistors had markings like resistors and printing numbers and letters on a 3D object was more problem than it was worth until a high volume innovation was realized as the number of parts multiplied over the years. Some later parts have both part number and color bars.

Therefore, I am surprised that my theory is wrong as you note the same part has different color bars. My theory is out the window.
 

#12

Joined Nov 30, 2010
18,224
Color can mean anything. At one corporation; blank, red, and green were used to indicate input offset voltage range on the op-amps.
At another corporation, a red dot showed that a 2N3055 passed the 70 volt test.
You own the company, you make it mean whatever you want it to mean.
 

ErnieM

Joined Apr 24, 2011
8,377
I'd be surprised and disappointed if they didn't test 100%.
I used to work for a shop that made transistors, literally diffusing in dopants, slicing up the wafer, attaching to the header, testing, perhaps burn in...

The semiconductor tester they used could test just about anything you wish in unde a second, then light a display to indicate which bin to place it so just dead parts could be separated from weak parts (perhaps sold under a different part number) from good ones; it ha 99 bins in all.
 

dl324

Joined Mar 30, 2015
16,845
I used to work for a shop that made transistors, literally diffusing in dopants, slicing up the wafer, attaching to the header, testing, perhaps burn in...
I worked for a semiconductor manufacturer too and they tested 100% of their parts. Even if yields are very high, you still need to test to make sure all parts were within spec.
 

WBahn

Joined Mar 31, 2012
29,978
There could be lots of reasons. The one that comes to the top of my head is that they binned a bunch of transistors based on various parameters, possibly at the request of a major customer, and the color codes indicate the bin (parameter ranges). They almost certainly binned up many more transistors than were actually shipped to that customer, so the rest were placed in general stock.

This was probably far more common in years past given how tight production tolerances are today.
 

Thread Starter

Willen

Joined Nov 13, 2015
333
I worked for a semiconductor manufacturer too and they tested 100% of their parts. Even if yields are very high, you still need to test to make sure all parts were within spec.
Probably the testing was automated process during manufacturing (hundreds of transistor within a minute), when transistors are moving through conveyor belt, lets say. Otherwise testing their millions of transistors' spec one by one manually would take millions of hours.
 

Thread Starter

Willen

Joined Nov 13, 2015
333
Hi KISS,
The link says they saw some transistors colored to classify according to their hFE. I tested same colored transistors (mine have 3 color bands) and have very different hFE. As all said maybe the bands are for manufactures' purpose, not for us. So all are being unable to decode my bands 'exactly'.
 

dl324

Joined Mar 30, 2015
16,845
Probably the testing was automated process during manufacturing (hundreds of transistor within a minute), when transistors are moving through conveyor belt, lets say. Otherwise testing their millions of transistors' spec one by one manually would take millions of hours.
They use automated testers. The cost for testing discrete devices used to be a significant fraction of the selling price.
 

GopherT

Joined Nov 23, 2012
8,009
Hi KISS,
The link says they saw some transistors colored to classify according to their hFE. I tested same colored transistors (mine have 3 color bands) and have very different hFE. As all said maybe the bands are for manufactures' purpose, not for us. So all are being unable to decode my bands 'exactly'.
Willen,

The bars are different than binning color ( usually dots, not bars). It might be a date of manufacture and it might be a gain ( possibly AC gain). The interesting thing for me is that some very only transistors had no model numbers at all, just color bars like resistors. Also interesting that I cannot find a single reference to color bars/stripes/lines on a transistor anywhere on the interwebs.

Cheers.
 

DickCappels

Joined Aug 21, 2008
10,153
The use of color codes suggests a human was somehow involved.

As an aside, I have perhaps 2000 transistors that match the description in the first post. Coincidence??
 

GopherT

Joined Nov 23, 2012
8,009
The use of color codes suggests a human was somehow involved.

As an aside, I have perhaps 2000 transistors that match the description in the first post. Coincodence??
Human involved? How?
The three color bands Willen and I are talking about are machine made.
They are not the one or two splotches of color made by a human hand as the result of binning.
 
Top