My Transistors, original or copy?

[Primetime]

I've a batch of confirmed original MJL4281A, the value is a bit higher compared to the above. It ranges between 12 - 14 nF, avaraging about 13nF.

[Primetime]

For my batch of 20 MJL4302A, they are quite consistent at about 11 nF.

[anatech]

Hi all,
I've now measured some new MJL21195 and MJL21196 transistors. PNP and NPN current gains are about 2:1 at low currents. I'm going to have to try this at higher currents.

MJL21195 beta averages about 100 and are very consistent, B-E capacitance is averaging 7.05nF at 10 KHz.

MJL21196 beta ranged from 38 to 62 averaging 52, B-E capacitance varied from 5.02nF to 5.81nF averaging 5.33 nF at the same 10 KHz.

Capacitance meter test voltage was 100 mV AC.

If On Semi could only get the newer devices to be as consistent as the MJW parts, we'd all be in heaven.

-Chris

[mitwrong]

Hi, Ladies and Gents:
I want to say a word on transistors, To determine a transistor is not depends on it's b-c, b-e, e-c internal capacitance. When a piece of wafer which is to make transistor, the wafer is about 3 inches in dia. after photo, etching, and layer diffusion, thus many of the transistors were borned.
The point is during diffussion process, the doping in the materials
N or P which N means extra electrons, and P means extra holes.
due to the impurities in these material can come out with different resistance and capacitance. Then with these data to determine the transistor is orginal or false product is not true.
The correct method to determine a transistor is to melt out the casing, ( plastic ) then check under microscope, there would be a lot no. etched on the wafer. this is the only way to identify the transistor. And also consult from the manufacturer about the lot no. and date code.
The date code is eg. 8923, that means this was manufacturer in Yr. 1989, 23 week of that yr.
I'been worked in semicon factory for many yrs. If any Ur question rise, don't hesitate to ask.




WE CONSIDER THE TRUTH OF A MATTER

[clem_o]

Originally posted by mitwrong

due to the impurities in these material can come out with different resistance and capacitance. Then with these data to determine the transistor is orginal or false product is not true.
The correct method to determine a transistor is to melt out the casing, ( plastic ) then check under microscope, there would be a lot no. etched on the wafer. this is the

Ni Hao!

What everyone hopes for is a simple and non-destructive method of determining fake from genuine transistors, as claimed by its top-mark.

I agree that semiconductor process isn't an exact nor extremely accurate science (I worked at a semicon too, some years ago), thus microprocessors from even the same batch exhibit differences in maximum clock speed.

However, to within a certain 'tolerance' same transistor types will measure the same capacitance, which although partly influenced by impurities etc, will ALL have the same essential geometric properties (i.e. die size, etc), which I *think* has a greater effect on capacitance. And fortunately, the manufacturers of the 'fake' transistors often use dies that are smaller than the originals (we assume, to minimize cost).

The end result is that fake transistor's B-E capacitance is often significantly lower than that of the originals.

OTOH it does seem there are some exceptions... but seems that this is for the significantly older devices... (i.e. 2SB600)

Cheers!

[mitwrong]

to clem_o
Ur info. some are true, some are false.
The false transistor usually they bought the fall out wafer, and do the packing. thus decrease their R&D cost. and not the smaller die to make the tr.
No matter bigger or smaller die, the manufacturing cost almost the same, only the failure rate is diff. the bigger the more, thus due to the diffusion control technique.

with the same diffusion process, the result would have been diff.
It depends on the control of the process, ie: oven temp, turing speed, etching process---- many factor will cause the defects.
may U agree with my opinon?

PS: I am not telling people to destructive, To find out the truth, U may sample size to destruct, this may worthwile in between truth and false. ( sacrface in return of valuable )



WE CONSIDER THE TRUTH, NOTHING BUT THE TRUTH

[anatech]

Hi mitwrong,
I normally recommend a visual inspection followed by DC beta and capacitance tests. Not perfect but it should work in most cases. There are always ways to get around the system.

-Chris

[clem_o]

Originally posted by mitwrong

The false transistor usually they bought the fall out wafer, and do the packing. thus decrease their R&D cost. and not the smaller die to make the tr.

mitwrong,

So they use "reject" dies? Oh my, that's definitely going to be a problem... Thanks for pointing this out as a possibility!

If some of the fakes are actually rejects from the SAME wafer, then wouldn't they have the same batch number anyway?

A question: what are the possible reasons for rejecting at wafer probe? What I can think of:

1. Gross failure (i.e. not at all functional)
2. Low Beta
3. Beta curve out of spec
4. Low ft

Did I miss anything?

Cheers

[anatech]

Hi Clem,
Yes.
The expensive part is attaching the die to a casing I would think. I don't see the point of this. Renumbering an existing part bought very cheaply makes far more sense.

Am I on the right track here?

-Chris

[clem_o]

Hi Chris,

Good point! Can't verify, either I wasn't listening or they never told me how much die-attach costs... :-)

Perhaps lump-summing everything into 'packaging' is the easier way to think about it (i.e. epoxy molding, die-attach, wire-bonding, plus the cost of materials - copper base, wire frame, epoxy); how does this compare to the cost of a transistor die???

Cheers

Clem