If you have a large difference in HFE between NPN and PNP output devices. will this have any ill effects such as distortion or more likley to oscillate etc?
For example. In my latest venture i have an amplifier with 4 pairs of output devices. the NPN's have an Hfe in the 10's and the PNP's have a Hfe in the high 80's.
Pretty large range. Now these devices came from the factory supposedly matched. i am assuming that they have matched these to whatever specs they prefer. A quick check with my little transistor checker shows that they are all grouped similarly.
So, im curious, with this sort of assymetry of Hfe, would this output stage be more prone to Oscillation? or have more distortion products then say the exact same set up with the Hfe's of NPN and PNP that were much closer to each other?
Understandable it may not be possible to get devices that close, but lets assume that we could. or at least get them with in 20 of each other. NPN's in the 50's and PNP's in the 60's for example. and of course each matched to like within there class.
Would an output stage as such be less oscilaltion prone or have less distortion or other ill effects??
Zc
For example. In my latest venture i have an amplifier with 4 pairs of output devices. the NPN's have an Hfe in the 10's and the PNP's have a Hfe in the high 80's.
Pretty large range. Now these devices came from the factory supposedly matched. i am assuming that they have matched these to whatever specs they prefer. A quick check with my little transistor checker shows that they are all grouped similarly.
So, im curious, with this sort of assymetry of Hfe, would this output stage be more prone to Oscillation? or have more distortion products then say the exact same set up with the Hfe's of NPN and PNP that were much closer to each other?
Understandable it may not be possible to get devices that close, but lets assume that we could. or at least get them with in 20 of each other. NPN's in the 50's and PNP's in the 60's for example. and of course each matched to like within there class.
Would an output stage as such be less oscilaltion prone or have less distortion or other ill effects??
Zc
Oscillations I don't know, could be. But they sure will increase distortion. With such a low Hfe the driver will be loaded heavier and increase distortion.
Are you SURE the NPN's have Hfe in the 10's?? Very unusual, even cheap power transistors (if they are genuine) should have > 50 Hfe.
Normally the PNP's have lower Hfe within a complementary pair.
What are the types?
Jan Didden
Are you SURE the NPN's have Hfe in the 10's?? Very unusual, even cheap power transistors (if they are genuine) should have > 50 Hfe.
Normally the PNP's have lower Hfe within a complementary pair.
What are the types?
Jan Didden
MJ150024 & 25's
These were the numbers from my little tester.
NPN - Hfe- 7, 8, 8, 12
PNP - Hfe- 89, 89, 90, 93
Now this was not at the current level they will be running at. I dont know how the MFG matches these devices. But the NPN's seem very low. so i was curious what ill effects they will have being so far off.
again this is what was supplied to me by the MFG's service dept.
Zc
These were the numbers from my little tester.
NPN - Hfe- 7, 8, 8, 12
PNP - Hfe- 89, 89, 90, 93
Now this was not at the current level they will be running at. I dont know how the MFG matches these devices. But the NPN's seem very low. so i was curious what ill effects they will have being so far off.
again this is what was supplied to me by the MFG's service dept.
Zc
Hi,
you are testing at the wrong current.
Output devices, if matched, should be tested at a current near where they operate.
I would aim for operational quiescent current, but others may say typical output currents (per device) should be used.
A range of currents from half Iq to half Ipk (curve tracing) may be a good way but I fear impossible to achieve.
Distortion performance may improve.
Oscillation, unlikely.
You will need to test at low voltage to minimise temperature changes during testing, unless you have access to a low duty pulse measurement facility. Low voltage will affect the accuracy and matching but it's usually all us amateurs can achieve.
you are testing at the wrong current.
Output devices, if matched, should be tested at a current near where they operate.
I would aim for operational quiescent current, but others may say typical output currents (per device) should be used.
A range of currents from half Iq to half Ipk (curve tracing) may be a good way but I fear impossible to achieve.
Distortion performance may improve.
Oscillation, unlikely.
You will need to test at low voltage to minimise temperature changes during testing, unless you have access to a low duty pulse measurement facility. Low voltage will affect the accuracy and matching but it's usually all us amateurs can achieve.
Lol, I just exerienced my first wrong connection problem when stuffing one of my Nuuk class A buffer boards a few minutes ago... Ran out of bc547 so I tried sticking a bc639 into the socket without proper orientation. Just got clicking sound from outputs... hey I'm stil a transistor noob... only learned 2 minutes ago about TO92A B and C. All great now! Now to wire the volume and balance controls between the 4 buffers (2 per channel with attenuator between stages).
My very first fully fledged solidstate hardware
My very first fully fledged solidstate hardware
Nordic said:
First thing I do when starting to layout a pcb or assemble something is to get copies of all the data sheets into the project binder and check each type before inserting/soldering. Fell on my face too many times. Also great when needing to replace something xx years later...
Jan Didden
Because i did not test at operating currents, i am assuming my numbers are wrong. I was using my little M3 transistor tester that measures Hfe at a low current.
However i did test 8 other devices at that same time and got readings more in the norm. 3 that were originals in the amp that survived and 5 new Onsemi sample parts.
The parts in question were supposedly factory matched from Classe. I now have the devices installed so i cannot retest them. But i am assuming that they are matched to whatever specs that were looking for.
The surviving transistors are no good to me. i wont trust them in a different amp so they become test pigs. I will use these to set up a test rig to measure Hfe at operating current etc.
But lets put that aside for a moment and lets pretend that we have an amplifier where the N side has matched devices in the 7-10 Hfe range and P devices in 80-90 Hfe range.
My first thought would be that the amp would be more prone to oscillation then an Amp where the N and P sides had Hfe's closer together. From comments here that doesnt seem to be the case.
My second thought was that the the amp may have some form of asymetrical waveform problems resulting in some form of distortion. or that possible one half of the wave form would clip sooner then the other.
How far apart can these Hfe's be before a problem is apparent?
If we have devices in the 10's how far off could the other side be before there is a gross error?
Thanks everyone for your input, theoretical questions like this help me understand things better.
Zc
However i did test 8 other devices at that same time and got readings more in the norm. 3 that were originals in the amp that survived and 5 new Onsemi sample parts.
The parts in question were supposedly factory matched from Classe. I now have the devices installed so i cannot retest them. But i am assuming that they are matched to whatever specs that were looking for.
The surviving transistors are no good to me. i wont trust them in a different amp so they become test pigs. I will use these to set up a test rig to measure Hfe at operating current etc.
But lets put that aside for a moment and lets pretend that we have an amplifier where the N side has matched devices in the 7-10 Hfe range and P devices in 80-90 Hfe range.
My first thought would be that the amp would be more prone to oscillation then an Amp where the N and P sides had Hfe's closer together. From comments here that doesnt seem to be the case.
My second thought was that the the amp may have some form of asymetrical waveform problems resulting in some form of distortion. or that possible one half of the wave form would clip sooner then the other.
How far apart can these Hfe's be before a problem is apparent?
If we have devices in the 10's how far off could the other side be before there is a gross error?
Thanks everyone for your input, theoretical questions like this help me understand things better.
Zc
Hi,
the driver on the low gain half will be supplying more current to satisfy the load on the output. i.e. the driver sees a lower impedance.
this will cause the driver to run nearer it's SOA limit.
The lower impedance may cause the open loop distortion to be higher and then the global is working harder to hide it.
the driver on the low gain half will be supplying more current to satisfy the load on the output. i.e. the driver sees a lower impedance.
this will cause the driver to run nearer it's SOA limit.
The lower impedance may cause the open loop distortion to be higher and then the global is working harder to hide it.
Hi Zero,
was there another thread that caused confusion?
output see the load.
drivers see the output stage as their load, ~=load*hFE.
predriver/VAS sees the drivers as their load. etc.
Hang a reactive load on the output and the effect ripples back through all the stages pushing each device towards destruction.
was there another thread that caused confusion?
output see the load.
drivers see the output stage as their load, ~=load*hFE.
predriver/VAS sees the drivers as their load. etc.
Hang a reactive load on the output and the effect ripples back through all the stages pushing each device towards destruction.
Hi ZeroCool,
Part of your concern with running the low gain devices depends on the amplifier topology you’re going to run: Full complimentary/symmetry, dual differential vs. single differential pair with an opposing constant-current source, or some variation of each will slightly alter the feedback’s responsiveness to loads and ultimately the distortion produced. Plus, it also depends on if you’re going to run a Darlington pair or triple-Darlington output. With a triple-Darlington output, you may have no worries as you will have sufficient gain to unload the preceding stage. I’d think a dual differential input may be a little more sensitive to gain/load variations, but a CCS is already using a different current sourcing system and should cope better.
If you’re going to use some other variation like zero-feedback or single-ended, I’d be more worried about closer matched devices (basically because you’re going to have all sorts of other problems anyways).
When I tested some MJ15022/23 devices with my M3, some of the 22’s did measure in the 7 – 10 range, but most fell in the 40 – 50 range. Your measurements do seem low, but I don’t know if changing your testing method to get the numbers you are looking for is the answer.
So if your M3 says hFE = 10, and the same device tested at 500mA says hFE = 30, it’s still the same device. The only thing that’s changed is your peace-of-mind and a little understanding of how the device will operate in in-situ.
My last thought would be that it depends on your ultimate goal for the amp. If the amp’s going to be your ultimate center-piece, I’d spend the extra bucks and get more devices to choose from. If it’s an experiment to validate a topology or construction method, or maybe a subwoofer amp (where some HF distortion is acceptable), then I wouldn’t worry.
Paul
Part of your concern with running the low gain devices depends on the amplifier topology you’re going to run: Full complimentary/symmetry, dual differential vs. single differential pair with an opposing constant-current source, or some variation of each will slightly alter the feedback’s responsiveness to loads and ultimately the distortion produced. Plus, it also depends on if you’re going to run a Darlington pair or triple-Darlington output. With a triple-Darlington output, you may have no worries as you will have sufficient gain to unload the preceding stage. I’d think a dual differential input may be a little more sensitive to gain/load variations, but a CCS is already using a different current sourcing system and should cope better.
If you’re going to use some other variation like zero-feedback or single-ended, I’d be more worried about closer matched devices (basically because you’re going to have all sorts of other problems anyways).
When I tested some MJ15022/23 devices with my M3, some of the 22’s did measure in the 7 – 10 range, but most fell in the 40 – 50 range. Your measurements do seem low, but I don’t know if changing your testing method to get the numbers you are looking for is the answer.
So if your M3 says hFE = 10, and the same device tested at 500mA says hFE = 30, it’s still the same device. The only thing that’s changed is your peace-of-mind and a little understanding of how the device will operate in in-situ.
My last thought would be that it depends on your ultimate goal for the amp. If the amp’s going to be your ultimate center-piece, I’d spend the extra bucks and get more devices to choose from. If it’s an experiment to validate a topology or construction method, or maybe a subwoofer amp (where some HF distortion is acceptable), then I wouldn’t worry.
Paul
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.