Today i finally decided to measure the static hfe of all my stock of TO3 cans MJ15003/4.
All my power transistor were bought in batches of about 50 over the time, only from mayor vendors (i.e. Farnell , etc.).
ALL are original Motorola/ON (Mex factory) , all never used in production.
I had about 120 BJTs to test, about half NPN and half PNP.
For the test, I had set a base current of about 3,5 mA , measuring the Ic on the 1 Ohm collector resistor. This gives usually 100-600 mA Ic, enough to make the transistors pretty warm with 10V collector supply. Measure was done always after 10 seconds of connection to the test circuit.
The results were quite surprisinng.
1) The old Motorolas had quite ALWAYS an hfe almost double than the modern ONsemi. Typical values of 220 for the -04 and 120 for the -03 for the Motorolas VS 150 for -04 and 60 for the -03 on the new ONsemi.
2)The -04 has CONSISTENTLY more than double the hfe than the -03 in the same batch. (batch is easily found because the transistor has the Year/Week prodution date on the case).
Inside the same batch the "variance" is quite low, around +-10%.
My conclusions:
1) You cant' assume that you can use any ONsemi transistor for repairing an amplifier using the old Motorolas, but you have to carefully select it.
2) When designing a circuit you cannot assume that the -04 is REALLY a complementary of the -03.
I would like to hear some opinion on the argument.
Thanks in advance
effebi
All my power transistor were bought in batches of about 50 over the time, only from mayor vendors (i.e. Farnell , etc.).
ALL are original Motorola/ON (Mex factory) , all never used in production.
I had about 120 BJTs to test, about half NPN and half PNP.
For the test, I had set a base current of about 3,5 mA , measuring the Ic on the 1 Ohm collector resistor. This gives usually 100-600 mA Ic, enough to make the transistors pretty warm with 10V collector supply. Measure was done always after 10 seconds of connection to the test circuit.
The results were quite surprisinng.
1) The old Motorolas had quite ALWAYS an hfe almost double than the modern ONsemi. Typical values of 220 for the -04 and 120 for the -03 for the Motorolas VS 150 for -04 and 60 for the -03 on the new ONsemi.
2)The -04 has CONSISTENTLY more than double the hfe than the -03 in the same batch. (batch is easily found because the transistor has the Year/Week prodution date on the case).
Inside the same batch the "variance" is quite low, around +-10%.
My conclusions:
1) You cant' assume that you can use any ONsemi transistor for repairing an amplifier using the old Motorolas, but you have to carefully select it.
2) When designing a circuit you cannot assume that the -04 is REALLY a complementary of the -03.
I would like to hear some opinion on the argument.
Thanks in advance
effebi
Interesting about the betas between MOT vs ON semi not sure what the conclusion is per the spec.
BTW I think the betas between PNP and NPN drops much closer as Ic max is approached ~ 8A,
Matching device betas on each rail (current sharing) is more important than PNP vs NPN where the driver beta differences also tend to equal them.
BTW I think the betas between PNP and NPN drops much closer as Ic max is approached ~ 8A,
Matching device betas on each rail (current sharing) is more important than PNP vs NPN where the driver beta differences also tend to equal them.
Current sharing is my main concern.
I have around in operation some amps with 10 pairs where some BJTs run noticeably hotter than the other ones, Emitter resistor notwithstanding.
BTW I have tested also other ab. 30 "used" devices, exactly the same results.
Good news: Only one "bad" new device, a 15004 with hfe = 18.
I have around in operation some amps with 10 pairs where some BJTs run noticeably hotter than the other ones, Emitter resistor notwithstanding.
BTW I have tested also other ab. 30 "used" devices, exactly the same results.
Good news: Only one "bad" new device, a 15004 with hfe = 18.
That's right. I usually would do the same, with 2 exceptions:
1- you are on-site and time constrained.
2- Customers wants to "spare".....
You may very well get the same variance between two old batches of Moto and two new batches of ON's. Nothing has really changed between Moto and ON - that kind of lot to lot variation is normal in power transistors. You'll probably find that Vbe at a given Ic is closer than the beta spread between the batches - and that's what really determines current sharing.
Yes temp is important! If the OP measures same Ic and Vce @ 10 seconds would be acceptable.
BTW beta is a more accurate term here than hfe
Surely that currently, Onsemi send the higher grade to its
best customers and the remaining end at Mouser or other
general public orientated resellers...
Of course. It did not want to be a high precision measurement. Otherwise, to test the Transistor at such currents you probably would have to properly set it on a heat sink and control temperature.
Also, as you say, it would be better to set a costant Ic since it influence very directly the temperature.
To make the thing easier I just did set a (pseudo) constant Ib via a simple resistor bias network. Then, of course Ic varies (almost) linearly with hfe.
Doing the test for all the 150 transistors in the same conditions anyway lets me group them and find out outiers ("bad ones").
But the spread that I have is very large and very ripetitive.
I had transistors coming from different batches from 1990 to 2010, but the trends that I have summerized in my first post remain always valid.
On a final note I want to say that I am not blaming Motorola (now ONsemi) because I always had good experience with their BJTs on the reliabilty side.
I am just trying to find out what is going on, and that two trends look to me surprising.
Well, it won't be a problem if they just let us know that. I mean if the would put a beta class (i.e. A,B,C), it wont't be a big problem to pay a premium price for them. I also wonder how many you have to buy to become a "corporate" customer. If you try to order 200 of them, you will see that most of times they deliver you in 2 or 3 shipments. It does not seem that they are selling (or at least stocking) so many of them.
Effi,
your method is completely flawed.
Take a pair of power BJTs.
A with hFE = 100 @ 25degC and = 150 @ 100degC
B with hFE = 110 @ 25degC and = 160 @ 100degC
test both with the same Ib and same Ta.
A will rise in temperature due to junction dissipation increasing die temp (Tj)
it will hit a particular Ic @ 10seconds. You could plot Ic vs time from t0s to t10s
Now test B with exactly the same start conditions.
Tjb will initially rise faster than Tja. After one second it will be passing more than 10% extra Ic, because it's hFE will have risen more than the A with the slightly lower hFE due to passing more current and dissipating more power.
After 2seconds Ica will have risen a bit further, but Icb starting from a higher Tj1b will rise even more than the increase in A, due to an exponential rise in warm hFE, as the Tj deviate from the starting conditions.
This divergence of delta T over the 10seconds will push the B higher hFE to higher Ic and higher dissipation and higher Tj with the result you will measure an very big difference in hFE because A & B have different delta T.
Your method exaggerates the small 10% hFE difference and could end up showing as +20%, or +30% or even much worse. The differences you have measured are fiction.
If you want to measure genuine differences in hFE you must change your method. Something that allows you to measure each sample at very similar Tj.
Even measuring the hFE cold, i.e after 10ms from start of Ic flowing, would be a far better test than after 10s.
your method is completely flawed.
Take a pair of power BJTs.
A with hFE = 100 @ 25degC and = 150 @ 100degC
B with hFE = 110 @ 25degC and = 160 @ 100degC
test both with the same Ib and same Ta.
A will rise in temperature due to junction dissipation increasing die temp (Tj)
it will hit a particular Ic @ 10seconds. You could plot Ic vs time from t0s to t10s
Now test B with exactly the same start conditions.
Tjb will initially rise faster than Tja. After one second it will be passing more than 10% extra Ic, because it's hFE will have risen more than the A with the slightly lower hFE due to passing more current and dissipating more power.
After 2seconds Ica will have risen a bit further, but Icb starting from a higher Tj1b will rise even more than the increase in A, due to an exponential rise in warm hFE, as the Tj deviate from the starting conditions.
This divergence of delta T over the 10seconds will push the B higher hFE to higher Ic and higher dissipation and higher Tj with the result you will measure an very big difference in hFE because A & B have different delta T.
Your method exaggerates the small 10% hFE difference and could end up showing as +20%, or +30% or even much worse. The differences you have measured are fiction.
If you want to measure genuine differences in hFE you must change your method. Something that allows you to measure each sample at very similar Tj.
Even measuring the hFE cold, i.e after 10ms from start of Ic flowing, would be a far better test than after 10s.
Last edited:
I don't agree with You. I have 140-140 pcs. of NJW0281/0302 from Mouser. I measured the Hfe, and 100% of them are within the specification, especially the 95% of the PNP-s are between 100 and 110. Really good...
Sajti
I understand your complain about the measuring method , you are surely right.
That was only a easy way to make a measure. 10s is a right time for an old man like me to turn power on, let stabilize the DVMs and read and write down.
But, please go back and read my first post.
What I (tried to) say is that *INSIDE A BATCH* the tolerance *IS* OK.
What is not explained (even by your correct teory) is why:
------To simplify the question to the extreme ( I know, there is a language ------problem, I am not flent in English)
1)Motorolas have ALWAYS about double beta than ONsemi
2)PNP have always about double beta than NPN
If the problem was in the measuring method I would obtain very noisy and uncorrelated data. That is not. My impression is that there is a technical difference (due to differences in production methods, materials, quality choice, this I cannot say).
That was only a easy way to make a measure. 10s is a right time for an old man like me to turn power on, let stabilize the DVMs and read and write down.
But, please go back and read my first post.
What I (tried to) say is that *INSIDE A BATCH* the tolerance *IS* OK.
What is not explained (even by your correct teory) is why:
------To simplify the question to the extreme ( I know, there is a language ------problem, I am not flent in English)
1)Motorolas have ALWAYS about double beta than ONsemi
2)PNP have always about double beta than NPN
If the problem was in the measuring method I would obtain very noisy and uncorrelated data. That is not. My impression is that there is a technical difference (due to differences in production methods, materials, quality choice, this I cannot say).
Hi
Measure hfe with multimeter just dosn't make any sense.....use curve tracer or build tester which inject in transistor short current impulses and has hold circuit for meter. Such circuit was published by elektor in eighties.....if someone wants I can post schematic. But next week because I'm on vacation....when I come back home. Btw. I also nocited higher hfe at older motorola devices. But on the other hand. If you have device with something like 50 of amplification will get over 20A at some 0.5A of drive.....who needs more...hehe
Best regards
Taj
Measure hfe with multimeter just dosn't make any sense.....use curve tracer or build tester which inject in transistor short current impulses and has hold circuit for meter. Such circuit was published by elektor in eighties.....if someone wants I can post schematic. But next week because I'm on vacation....when I come back home. Btw. I also nocited higher hfe at older motorola devices. But on the other hand. If you have device with something like 50 of amplification will get over 20A at some 0.5A of drive.....who needs more...hehe
Best regards
Taj
Answering a post without reading it does not make any sense.
As I have stated from the beginning I did *NOT* use a multimeter.
Even without getting at that, show me a multimeter that sets a 500 mA Ic.
Then I agree that my method does not resent the perfection that here seems to be necessary for any thing done, but I do not need to be patronized for that, thank you.
My intent was only to share my experience on the subject.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.