Hi Andy,
A pulsed SOA is every bit as destructive as steady state. For steady state you must compensate more for the case temperature. You will have more variables to deal with regarding heatsinking.
The second breakdown occurs when local hot spots are created in the die and that area goes into thermal runaway. It happens so quickly that the case is not a strong factor in getting rid of the heat. A pulse test is also much easier on your power supply design.
-Chris
A pulsed SOA is every bit as destructive as steady state. For steady state you must compensate more for the case temperature. You will have more variables to deal with regarding heatsinking.
The second breakdown occurs when local hot spots are created in the die and that area goes into thermal runaway. It happens so quickly that the case is not a strong factor in getting rid of the heat. A pulse test is also much easier on your power supply design.
-Chris
vectorplane said:
Clem:
My offer still holds, in case your part search is unsuccessful. I can ship at no charge (from my employer's mailroom
I'll consider it a go for that project using a PIC to test transistors. I'll put together a schematic and go from there. Are you ok with programming some of the new dsPIC devices? Or should we rather stick with 16F series parts, for low cost?
Andy,
Its a mighty generous offer! I'll do one last round this week if I get a chance, and let you know.
Lets stick to the 16F family - availability here is questionable for the other families, though 18F seems to be trickling in. The 'net also has tons of home-brew programing hardware and software for the medium-range family, so in case anyone else is interested in building whatever this project produces, it'll be easier to duplicate...
Cheers!
Clem
The small dice used in the counterfeits have much less contact area with the case; so there will actually be thermal differences, with the counterfeit having a higher j-c thermal resistance.
It's also quite possible that the die attach on the counterfeit often may not even include the entire die surface, so the difference would be even greater.
I'd suspect that the counterfeits are so inferior to the genuine, that you don't really have to crowd the SOA envelope that closely. Thermal mass of the counterfeit dice appears to be less than half that of the originals. If you do decide to use pulsed, single shot should be adequate.
It's also quite possible that the die attach on the counterfeit often may not even include the entire die surface, so the difference would be even greater.
I'd suspect that the counterfeits are so inferior to the genuine, that you don't really have to crowd the SOA envelope that closely. Thermal mass of the counterfeit dice appears to be less than half that of the originals. If you do decide to use pulsed, single shot should be adequate.
If this is so, we might be able to get away with simply allowing the transistor in question dissipate some power, measure the case temp and compare it with an estimate of the junction temp (based on the forward conduction voltage). If there's a fairly large temp difference, it would point to a small or weak die-case bond. Could be very useful as this would be non-destructive...
Comments?
Cheers,
Clem
Comments?
Cheers,
Clem
Hi Clem,
That would be highly variable. It depends on mounting pressure, thermal efficiency of the heatsink and possibly ambient temperature. For a non-destructive test I am leaning towards capacitance and beta testing.
Keep in mind that a poor fake (as many are) may fail the simplest of tests that are well within the SOA of a normal part. You may still end up with a dead part.
-Chris
That would be highly variable. It depends on mounting pressure, thermal efficiency of the heatsink and possibly ambient temperature. For a non-destructive test I am leaning towards capacitance and beta testing.
Keep in mind that a poor fake (as many are) may fail the simplest of tests that are well within the SOA of a normal part. You may still end up with a dead part.
-Chris
I think thermal properties of counterfeit transistors are probably not consistent enough to allow reliable screening.
Some fakes are actually genuine transistors (cheap ones) lika a 2N3055, relabelled as something else.
Moreover, the instrumentation necessary to accurately measure thermal performance is not trivial. You can't just scotch-tape a thermometer's bulb to the transistor case, and read off values. The size, shape and orientation of your heat sink, the thermal resistance case-to-sink, mounting method and force, air circulation in the room, the calculations begin to hurt the brain.
Your setup may be different than mine. Your ambient temp may be 35 C and mine may be 18C.
We need to go back to the basic question: what is it that makes expensive transistors expensive.
1. Exceptional SOA. An SOA curve that bulges out in the middle of the graph, where others cave in.
2. Sustained hfe up to high currents.
Device signature, like capacitance, resistance, Vbe, etc, is all nice to measure, but there is a chance you may get some false negatives this way (i.e. the fake passes). Qualifying a device by signature typically requires mesauring several parameters, not just one, some of which you may not have published values for.
I would avoid generalizing that these tests are deterministic. Besides, you may need a reference (good) device to compare the DUT readings against. You may need MANY reference devices, if part-to-part variability in good devices is high.
I'd tend to think that if a fake passed [1] and [2] above, it would then not ba a fake. In 14 fakes I've come across , I've never met one that passed them. Because it is an expensive process to make devices that handle these stresses.
Plus, you don't need reference devices; you can use published specs.
Now for the tough part.
A PIC could test signature parameters at low powers. But if you want it to test SOA, you'll need an external power supply capable of over 100V and 5 amps or so. Plus a Buck converter (controlled by the PIC) to lower this voltage for some tests. Not impossible, just not very portable....
Unless...
You use a Li-ion (1Ah) battery to power a boost converter which charges a large eletrolytic cap (slowly) to some programmed voltage. Then use the charge in that capacitor for the SOA tests.
Photographic flash units use this approach all the time.
Andy
Some fakes are actually genuine transistors (cheap ones) lika a 2N3055, relabelled as something else.
Moreover, the instrumentation necessary to accurately measure thermal performance is not trivial. You can't just scotch-tape a thermometer's bulb to the transistor case, and read off values. The size, shape and orientation of your heat sink, the thermal resistance case-to-sink, mounting method and force, air circulation in the room, the calculations begin to hurt the brain.
Your setup may be different than mine. Your ambient temp may be 35 C and mine may be 18C.
We need to go back to the basic question: what is it that makes expensive transistors expensive.
1. Exceptional SOA. An SOA curve that bulges out in the middle of the graph, where others cave in.
2. Sustained hfe up to high currents.
Device signature, like capacitance, resistance, Vbe, etc, is all nice to measure, but there is a chance you may get some false negatives this way (i.e. the fake passes). Qualifying a device by signature typically requires mesauring several parameters, not just one, some of which you may not have published values for.
I would avoid generalizing that these tests are deterministic. Besides, you may need a reference (good) device to compare the DUT readings against. You may need MANY reference devices, if part-to-part variability in good devices is high.
I'd tend to think that if a fake passed [1] and [2] above, it would then not ba a fake. In 14 fakes I've come across , I've never met one that passed them. Because it is an expensive process to make devices that handle these stresses.
Plus, you don't need reference devices; you can use published specs.
Now for the tough part.
A PIC could test signature parameters at low powers. But if you want it to test SOA, you'll need an external power supply capable of over 100V and 5 amps or so. Plus a Buck converter (controlled by the PIC) to lower this voltage for some tests. Not impossible, just not very portable....
Unless...
You use a Li-ion (1Ah) battery to power a boost converter which charges a large eletrolytic cap (slowly) to some programmed voltage. Then use the charge in that capacitor for the SOA tests.
Photographic flash units use this approach all the time.
Andy
testing, testing...
Hi,
I think I read somewhere that someone had successfully & reliably differentiated fakes from real using the capacitance of the CB & EB junctions. Basically the capacitance of the fakes is much lower due to the smaller die...
Wish I could remember where I read about it...
Stuart
Hi,
I think I read somewhere that someone had successfully & reliably differentiated fakes from real using the capacitance of the CB & EB junctions. Basically the capacitance of the fakes is much lower due to the smaller die...
Wish I could remember where I read about it...
Stuart
You betch ya.
I once tested a bunch of (about 15) 2N3055's, trying to find the one with highest hfe at rated current, for a circuit.
They all hovered around 4, some going to 5. There was absolutely no way any of these stubborn parts was going to give me the min. hfe of 8 that I needed.
They were impeccably consistent in their lack of performance.
Andy
Choice of technique is governed by test equipment; available, build or buy? And destructive vs. non-destructive. .
SOA requires high currents, moderately high voltages, is destructive and a very good test; and fakes are removed from the stream of commerce forever. But it requires a significant tester. Beta at high currents could be checked using many of the SOA tester's capabilities as well.
Capacitance can be spoofed by using a large cheap die; low beta, low Vce, but high capacitance.
Vce might be helpful. Ramp voltage up to 1 mA leakage, capture and shut down, but Vce can be spoofed by using a small, high voltage chip.
So I guess I've meandered around to preferring either SOA or the combination of capacitance and Vce which do not require high currents.
SOA requires high currents, moderately high voltages, is destructive and a very good test; and fakes are removed from the stream of commerce forever.
But it requires a significant tester. Beta at high currents could be checked using many of the SOA tester's capabilities as well. Capacitance can be spoofed by using a large cheap die; low beta, low Vce, but high capacitance.
Vce might be helpful. Ramp voltage up to 1 mA leakage, capture and shut down, but Vce can be spoofed by using a small, high voltage chip.
So I guess I've meandered around to preferring either SOA or the combination of capacitance and Vce which do not require high currents.
a percentage of real devices?... Ouch!
Of course you don't test close to the axes. (hi V - lo I ). Any cheap transistor does well there.
The expensive parts are expensive because of what they do in the MIDDLE of the SOA curve. Medium voltage/medium current.
I think the application implied here is Class B audio amps. The load lines for such circuits come very close to the SOA at moderate voltage and current levels.
You test 3-5 points in the middle of the curve, using perhaps the 10ms and the DC traces, short pulse test.
Come to think of it, you don't even need switching regulators to scale the power to the DUT, simple linears will do, because they only operate for a second or two, and don't have time to get hot.
Andy
Ok, so what have we got here while I was in dreamland?
Suggest then capacitance test as first screen, followed by pulsed SOA test in the 'middle' of the curve if needed. Should be ok, hopefuly many of the fakes will be weeded out by the first test. Purchase can then be made on a sample unit for further (possibly destructive) SOA testing...
Waiting for the interface circuits. Andy, send me an e-mail outlining what you envision the PIC interface requirements to be, when you have time!
Cheers!!
Clem
Suggest then capacitance test as first screen, followed by pulsed SOA test in the 'middle' of the curve if needed. Should be ok, hopefuly many of the fakes will be weeded out by the first test. Purchase can then be made on a sample unit for further (possibly destructive) SOA testing...
Waiting for the interface circuits. Andy, send me an e-mail outlining what you envision the PIC interface requirements to be, when you have time!
Cheers!!
Clem
Hi Chris,
Yes, noted re SOA, this would be the 'last' test, on the assumption that one buys the part and brings it home since it passed the first (set) of tests...
Seems it would be nicer to interface the tester to a notebook pc, if we are going to be testing Beta as well (I assume we'd want to see some sort of a curve right?). Then compare against manufacturer's published specs (or information gathered over several users, as I wouldn't know how much variation to expect!)
Cheers!!
Clem
Yes, noted re SOA, this would be the 'last' test, on the assumption that one buys the part and brings it home since it passed the first (set) of tests...
Seems it would be nicer to interface the tester to a notebook pc, if we are going to be testing Beta as well (I assume we'd want to see some sort of a curve right?). Then compare against manufacturer's published specs (or information gathered over several users, as I wouldn't know how much variation to expect!)
Cheers!!
Clem
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.