does early effect shift much within a device family?

[unclejed613]

my question is this: if i were to test several transistors that are the same part number, for instance a 2SD2390, but having different beta (the published Hfe range for a 2SD2390 is 5000-30000), will the early voltage a)remain about the same, or b) change radically with the beta?

[Lumba Ogir]

uncle,
the question is not worded properly. The Early effect or base-width modulation occurs due to the variation of the base-to-collector voltage. As the base width narrows (in extreme cases can become zero), the current gain rises and thereby the collector current. A lower Vce and beta reduces the Early effect, which is equal to beta*Early voltage product.

[unclejed613]

so i guess my question would be, would the Early voltage remain consistent between samples of the same transistor type? what i'm getting at is, could i calculate the Early voltage for a set of samples of a 2SD2390 and use that data to find out whether one of the samples is NOT a 2SD2390?

[Lumba Ogir]

uncle,
the Early voltage occurs dynamically, it cannot be derived from electrical properties nor can it be measured as a voltage. The Early effect gets smaller with decreasing Vce and beta.
Why are you especially worrying about this profane phenomenon? I`d just use those nice devices.

[Bob Cordell]

Originally Posted by unclejed613

so i guess my question would be, would the Early voltage remain consistent between samples of the same transistor type? what i'm getting at is, could i calculate the Early voltage for a set of samples of a 2SD2390 and use that data to find out whether one of the samples is NOT a 2SD2390?

Hi Unclejed. I think your question is reasonable and I think I understand it. First and foresmost, it is important to distinguish between Early effect and Early voltage when we are talking. Second, we need to recognize that Early effect is responsible for the output resistance of a transistor. The Early effect is a modulation of transistor beta by Vcb. As Vcb increases, the depletion region of the base-collector junction increases, encroaching on the base region and thinning the base. This increases beta.

An important Figure of Merit (FOM) for transistor families is the product of Beta and Early voltage VA. A transistor with beta of 100 and VA of 100V will have a FOM of 10,000V.

What you are asking, I think, is whether within a given family, is VA relatively constant in spite of Beta changes, or is FOM relatively constant in spite of Beta changes.

Moreover, I suspect you are wondering if different VA or FOM would indicate that a transistor was a fake.

I am not a device guy, but if I had to guess, I would think that a rogue transistor would more likely have a different FOM than a different VA. By extension, this would mean that if the real transistor and the fake had the same Beta, the lower-performing transistor would have a smaller VA.

I must admit here that I am on thin ice, and I could be totally wrong.

BTW, FOM is what largely governs output resistance of a transistor when connected as a cascode. High FOM is generally beneficial to the operation of a VAS.

Cheers,
Bob

[unclejed613]

yes that's exactly what i was trying to accomplish. i was wanting to come up with a somewhat simple way of detecting fakes. the company i work for gets inundated with them on a daily basis, and there is a huge amount of corporate inertia in their parts procurement process. we have several service facilities and the cost of using fakes is astronomical when rework and scrapping items as non-repairable is taken into account. in about 6 months of bulldogging this problem, very little has changed. the same knockoff parts come in on a regular basis. i have simple tools at my disposal for doing an incoming inspection. i measure Vbe forward and reverse, Ccb and Cce, and leakage current Iceo at 100 and 200V. those measurements plus a good visual and dimensional inspection usually is sufficient i have (after inspecting a LOT of known original parts) a rather large amount of data on mold marks, date coding methods, etc... that gives me about 90% correct answers on whether a device is genuine or not. it's at least a large portion of the other 10% i'd like to whittle down. i was thinking if there was a parameter that could be measured with simple equipment with a few well-placed data points, it would be something all of our service facilities could use with a minimum of training.

[Andrew Eckhardt]

Don't these fakers like to sell little dice for nominal price? Wouldn't it be possible to do a full DC beta sweep up to rated current and look for devices that have funny curves that crap out at low current? A number might be easy to fake but a curve wouldn't. Or for that matter just pick an arbitrarily high test current for a one point check. If it takes the power and it's a fake, it's a good fake.

[unclejed613]

that might be a good idea, i'll have to give it some thought. my goal is nondestructive testing as much as possible. our company has hired a testing company that specializes in detecting fakes, but that process takes time. if a fake is detected at the repair facility, i would like it if it survived long enough to get sent to the verifier. the purpose of having the other company do the verification is so that we can tell the parts supplier that they better get their act together. the purpose of getting the devices tested at the facility level is so that we can quickly shift gears and enact some workarounds to get the proper parts without customers wondering where their hifi's are.

[Andrew Eckhardt]

I think my Freudian slip was on to something. A real good fake might measure close in current gain but not have the SOA. You could try a repetitive pulse power test, something you could see on the scope to catch a die heating up that's too small. Maybe a jig that forced short, fixed, high power pulses could allow you to watch Vbe shrinking as the die heats up. Maybe it would be easier to drive fixed voltage into the base and watch collector current curve upward. You might be able to derive junction temp and maybe get some idea of relative die area. Average power need not be too high. You could stay well within ratings. I guess that's starting to get kinda complicated.

[unclejed613]

how DO they calculate the SOA charts? especially the 2nd breakdown region?????

i suppose if you're the device manufacturer, you can afford to break a few eggs to find out....