Possibly fake Sanken output trannies (PIC)

[capslock]

The Sanken 2SC2987 seemed to be sold out worldwide for a while. Only Mosspec and WEC "second sources" were available, but these should not really be called second sources because they didn't meet crucial parameters such as cob and fT.

Well, they seem to be back. I have attached a picture of what I just put my hands on. The parts look genuine enough to me. The only puzzling thing is that there is no gain group which all Sanken transistors had that I purchased in the last 3-4 years. Well, maybe they are unselected low cost parts.

I will measure cob and hFE and compare to older parts I still have. Should be able to post my findings tomorrow.

Regards,

Eric

[capslock]

Sorry, scanner PC already shut down. Will do a new scan tomorrow.

[capslock]

A1303 last year / this year. Old part has silver print and typical marking of two digit number (73) and gain group (O). Bottom center mold hole bearks SK mark.

New part is laser engraved or etched. Not clear whether marking O3A specifies gain group.

Back sides, old/new:
Back plate is smooth and flat without sanding marks. New part was sanded before molding.

[capslock]

now the front page pic

[capslock]

back side:

[capslock]

2SC2837, old and new.

Old part is laser engraved on rough rectangle. Code is F 0 27. Unclear wheter 0 actually means O, i.e. the gain group. No letters in the mold holes.

New part is laser engraved with smaller print. Writing is weaker , and when one looks more closely, it seems to be double printed with a diagonal offset. Either it really was printed twice, or the optics of the laser were really badly adjusted. Date code 0202 could or could not contain gain group.

[capslock]

Now for the C2837 back sides:

Old part was not sanded. Remaining horizontal structure may stem from rolling/pressing the original metal sheet. This is seen in some brand transistors.

New part was sanded before moulding.

[capslock]

Just for clarification: the A1303 is not the complemetary of the C2837. I just used it because I couldn't find my A1186s from last year.

[capslock]

h_FE as measured with a multimeter was consistent with spec sheets. To really tell genuine parts, one would need to do pulsed measurements at various collector currents up to 5 or 10 A.

A more easily accessible figure is the base collector capacitance. This is usually specified at some collector-emitter voltage of 10 - 80V and some frequency of 1 - 10 MHz.

With a multimeter, you measure without a CE voltage and at a few hundred Hz. However, this number is closely related to the spec values.

For genuine (silver print) Sanken parts (which are specified at 80 V and 1 MHz) and my multimeter, there is a factor of 3.5 - 4. Example

pair spec pF measurement
A1216/C2922 500/250 1620/945
A1303/C3284 400/200 1460/790

Another interesting thing to note is that for Sanken and other brand items, this capacitance varies +/-5% for parts with the same date stamp and maybe +/- 10% for parts from different lots. Unlike h_FE, the capacitance is determined by the geometry of the mask layout, and not by how well the diffusion process went for this wafer.

[capslock]

Last year's C2837 measured around 250-265 pF, which is a factor of 4.1 to the spec of 60 pF, so this is within the expected range.

The new parts are typically at 400-500 pF, two out of the ten shipped are at 260 and 275 pF, i.e. comparable to the old shipment. Another two have 703 and 786 pF. So on average, the capacitance is twice as high, and there is a really untypically high variation.

Just for comparison, two unreal second source C2837 from WEC measured 406 and 423 pF, i.e. the same as the new SK parts.

=> I would be tempted to assume that the C2837 parts I received are fakes.



The A1303 parts I received in the same shipment all measure 428 to 512 pF. This +/- 10% variation is a little unusual for a lot with the same date stamp, but much much closer than the C2837 parts. The really surprising thing is that this is virtually the spec value at 80 V and only a third of what I measured on parts with silver printing that look 100% genuine to me.

=> Either Sanken has improved the process dramatically (without changing the data sheet) or these are also fakes with a much smaller chip.