I'm looking for comments and feedback about the following idea:
What are the possibilities of using power BJT's closer to their measured Vce (or Vceo) breakdown voltage, but beyond their rated breakdown voltages? Naturally, they have to be used within their other stated parameters - namely their SOA. What are some of the potential pit-falls or concerns I should be aware of?
Here's how I came upon this:
I designed and built a DC junction breakdown tester (I could post a schematic and/or design information if there is interest) so I could measure the specs of unknown devices. It measures at either 25 or 100mA with a breakdown potential at either 160 or 320VDC. I then began measuring lots of parts. While measuring power transistors, there was no surprise to see that many parts measured within 10 to 20% of their rated voltage. A couple percent of devices unfortunately measured below their rated voltages too. But a lot of power BJT's measured significantly higher breakdown voltages than their ratings. Some were as high as 250% above their ratings! Plus, when I added some resistance from base to emitter, the breakdown voltage naturally increased. I was also surprised that when I heated the device, approaching 80C, the breakdown voltage further increased slightly.
This is certainly no surprise given how semiconductor manufacturers measure and rate their devices offering a spectrum of similar parts with the higher breakdown devices costing more. This is also why the data sheets state that some devices' parameters are listed as "guaranteed minimum..." with their actual voltages being some margin of error higher to compensate for real-world variations in manufacturing.
Here's some examples of parts I measured vs. their measured breakdowns:
TIP35C/TIP36C - rated 100Vceo, measured at up to 210V
MJE15030/MJE15031 - rated 150Vceo, measured at up to 220V
MJH11021/MJH11022 - rated 250Vceo, measured at up to >320V
And the list goes on.
My thoughts aren't to build monster amps and push the devices right to their hairy edge of breakdown, but to build amps with rails closer to the devices' ratings knowing there is a margin of error beyond its rating. It would also allow a wider range devices to work as a cross. Still, when using TIP35/36C output devices, I could use +/-50V and still be safe instead of +/-35, like I've done in the past.
I think I'll experiment with some devices and really push them to see how they perform.
I'd appreciate your comments, thoughts, experience, and suggestions.
Thanks,
Paul
What are the possibilities of using power BJT's closer to their measured Vce (or Vceo) breakdown voltage, but beyond their rated breakdown voltages? Naturally, they have to be used within their other stated parameters - namely their SOA. What are some of the potential pit-falls or concerns I should be aware of?
Here's how I came upon this:
I designed and built a DC junction breakdown tester (I could post a schematic and/or design information if there is interest) so I could measure the specs of unknown devices. It measures at either 25 or 100mA with a breakdown potential at either 160 or 320VDC. I then began measuring lots of parts. While measuring power transistors, there was no surprise to see that many parts measured within 10 to 20% of their rated voltage. A couple percent of devices unfortunately measured below their rated voltages too. But a lot of power BJT's measured significantly higher breakdown voltages than their ratings. Some were as high as 250% above their ratings! Plus, when I added some resistance from base to emitter, the breakdown voltage naturally increased. I was also surprised that when I heated the device, approaching 80C, the breakdown voltage further increased slightly.
This is certainly no surprise given how semiconductor manufacturers measure and rate their devices offering a spectrum of similar parts with the higher breakdown devices costing more. This is also why the data sheets state that some devices' parameters are listed as "guaranteed minimum..." with their actual voltages being some margin of error higher to compensate for real-world variations in manufacturing.
Here's some examples of parts I measured vs. their measured breakdowns:
TIP35C/TIP36C - rated 100Vceo, measured at up to 210V
MJE15030/MJE15031 - rated 150Vceo, measured at up to 220V
MJH11021/MJH11022 - rated 250Vceo, measured at up to >320V
And the list goes on.
My thoughts aren't to build monster amps and push the devices right to their hairy edge of breakdown, but to build amps with rails closer to the devices' ratings knowing there is a margin of error beyond its rating. It would also allow a wider range devices to work as a cross. Still, when using TIP35/36C output devices, I could use +/-50V and still be safe instead of +/-35, like I've done in the past.
I think I'll experiment with some devices and really push them to see how they perform.
I'd appreciate your comments, thoughts, experience, and suggestions.
Thanks,
Paul
Remember that the secondary-breakdown curves on the TIP35/36 start folding back above 30V, the MJH11021/22 fold back above 40V.
Last edited:
@ djk
No. I measured Vceo and Vce with the base shorted to the emitter. Would you mind explaining what the significance of SOA curves "folding back?" It's been quite a few years since my Advanced Linear class in college. LOL Looking at the SOA curves on the data sheets, I would naturally extrapolate the acceptable current at the higher Vce making sure I'm under the curve. And, for output devices, parallel them as necessary to ensure dependable operation even under worst-case reactive loading.
@ CBS240
Regarding the % of fails, I would only use devices beyond their rated voltage after I hand-measured each one and new what it could handle. Sorry if I wasn't clear about that. But just like hand-matching paralleled output devices or differential pairs for hfe and Vbe, I would also hand-measure the actual Vceo breakdown to ensure a good margin-of-error above my desired rails.
Regarding selecting the proper device for the job, that too would be the case for an OEM or new design - which I've done many times. The specific devices I listed I already have a good number of in stock. I'm just a DIY'er doing what I can, with what I have, and like most of us, always trying to push the boundaries of quality, performance, creativity, and function. I'm also trying to learn, help others, and get my hands dirty building audio gear (and robotics, but that's for a different forum).
No. I measured Vceo and Vce with the base shorted to the emitter. Would you mind explaining what the significance of SOA curves "folding back?" It's been quite a few years since my Advanced Linear class in college. LOL Looking at the SOA curves on the data sheets, I would naturally extrapolate the acceptable current at the higher Vce making sure I'm under the curve. And, for output devices, parallel them as necessary to ensure dependable operation even under worst-case reactive loading.
@ CBS240
Regarding the % of fails, I would only use devices beyond their rated voltage after I hand-measured each one and new what it could handle. Sorry if I wasn't clear about that. But just like hand-matching paralleled output devices or differential pairs for hfe and Vbe, I would also hand-measure the actual Vceo breakdown to ensure a good margin-of-error above my desired rails.
Regarding selecting the proper device for the job, that too would be the case for an OEM or new design - which I've done many times. The specific devices I listed I already have a good number of in stock. I'm just a DIY'er doing what I can, with what I have, and like most of us, always trying to push the boundaries of quality, performance, creativity, and function. I'm also trying to learn, help others, and get my hands dirty building audio gear (and robotics, but that's for a different forum).
I don't think that would work. SOA has to do with current bunching on the die at a specific actual Vce. The fact that a particular device would survive a higher-than-advertised Vce doesn't increase the SOA.
jan didden
@ janneman
Understood. Naturally at higher Vce the current would drop by the square of the increased voltage. Is this what you meant? It's why I mentioned the need for additional paralleled devices for the same load.
Understood. Naturally at higher Vce the current would drop by the square of the increased voltage. Is this what you meant? It's why I mentioned the need for additional paralleled devices for the same load.
"Would you mind explaining what the significance of SOA curves "folding back?" "
MJH11021 is 150W at 40V but only 40W at 100V, the power rating starts to fold-back above 40V.
As janneman points out the hot-spots on the die get worse at higher voltages.
Vceo is kind of meaningless in an audio amplifier when it is forward-biased under no-signal conditions.
MJH11021 is 150W at 40V but only 40W at 100V, the power rating starts to fold-back above 40V.
As janneman points out the hot-spots on the die get worse at higher voltages.
Vceo is kind of meaningless in an audio amplifier when it is forward-biased under no-signal conditions.
well ... here is my 2 cents
as about small devices ....lets say in a line of amplifier from input till drivers it normally pays to run semis as close to rated voltage .. most of them perform the best when working on the edge
seen goldmund amps running 100 v semis to 93 volts seen hardman cardon running drivers also too close to the maximum and since power wasn't enough use 2 of them in parallel to achieve the target .
Point is that in DC conditions you might get away with it ,... If the duty is steady and the conditions stabilized then you also might get away with it .
Point also is that in audio amplifiers there is way too many variables that actually decrease SOA of the all project ...number one variable is abuse .....
Also when it comes to long run you need also to take in mind thermals ,,,seen many of these amps with thermal/pcb/soldering issues when the choice is to run semis hot. Add to this the stabilization issues when semis are in parallel .
Finally .... to my understanding high quality audio amp ( with bjt class AB ) transistors can only be made with a couple of transistors only ( preferably sziklai ha ha ha ) increasing the number of outputs to get more power will degrade the quality ... to a small but existing ratio ......
Kind regards
sakis
as about small devices ....lets say in a line of amplifier from input till drivers it normally pays to run semis as close to rated voltage .. most of them perform the best when working on the edge
seen goldmund amps running 100 v semis to 93 volts seen hardman cardon running drivers also too close to the maximum and since power wasn't enough use 2 of them in parallel to achieve the target .
Point is that in DC conditions you might get away with it ,... If the duty is steady and the conditions stabilized then you also might get away with it .
Point also is that in audio amplifiers there is way too many variables that actually decrease SOA of the all project ...number one variable is abuse .....
Also when it comes to long run you need also to take in mind thermals ,,,seen many of these amps with thermal/pcb/soldering issues when the choice is to run semis hot. Add to this the stabilization issues when semis are in parallel .
Finally .... to my understanding high quality audio amp ( with bjt class AB ) transistors can only be made with a couple of transistors only ( preferably sziklai ha ha ha ) increasing the number of outputs to get more power will degrade the quality ... to a small but existing ratio ......
Kind regards
sakis
No this is incorrect. The fact that many manufacturers do that has to do with economics not with better performance. Most semi parameters deteriorate with higher voltage. Datasheets show that.
jan
I think it is just the opposite, with possibly the exception of input C everything else improves - bandwidth, distortion, damping factor, etc.
jan
The SOA is a property of the particular transistor used, not of anything else in the circuit or amp or 'project' whatever that is.
Of course you can exceed the transistor SOA, for instance by heavy, inductive loads like speakers with xover networks.
SOA says that at, for instance 80V Vce, your transistor can only deliver 1A (80W dissipation) even when the transistor is overall specced at 150W. With actual speaker loads which have a phase shift between load current and load voltage, you can have the top NPN transistor supplying current while Vout is negative! That means Vce is more than the pos suppply voltage, and SOA can be exceeded and poof again...
jan
Last edited:
that's OK Jan
i totally disagree with your sayings this is just my opinion ....
example 1
in an amplifier working with 35+35 rails ltp can be made with BC 556 , 2n5401 and MPSA 93 in most of the cases it will work far better with 556 obviously the reason is the one mentioned
example 2
Design amplifier with 6 outputs in a class AB circuit involves a wider pcb there distribution of power/ decoupling /ground /driving /thermals and capacitance will actually make the amplifier perform worst sound wise when compared to the same circuit if made with only a couple of transistors .....Since the target is the power obviously it can not be done otherwise ...
thats my opinion ...thank you for sharing yours
Kind regards
sakis
i totally disagree with your sayings this is just my opinion ....
example 1
in an amplifier working with 35+35 rails ltp can be made with BC 556 , 2n5401 and MPSA 93 in most of the cases it will work far better with 556 obviously the reason is the one mentioned
example 2
Design amplifier with 6 outputs in a class AB circuit involves a wider pcb there distribution of power/ decoupling /ground /driving /thermals and capacitance will actually make the amplifier perform worst sound wise when compared to the same circuit if made with only a couple of transistors .....Since the target is the power obviously it can not be done otherwise ...
thats my opinion ...thank you for sharing yours
Kind regards
sakis
I did not share my opinion; I try to base my posts on facts rather than opinions.
Opinions can vary all over the place and you never know what makes sense and what not.
Then again, I should remind myself we live in a post-fact society 😉
jan
Thank you Jan and jdk. I think we're all saying the same thing. As in my first post, SOA, thermal deration, etc. all must be carefully accounted for in the amp's design - and for my post, particularly the power outputs. I've built a number of power amps over the years, including a very nice 720W (when bridged into 4 ohms) car amplifier (including the 1.5KW DC/DC converter!), and worked for an amplifier OEM for a number of years - all with very good success.
I've attached the SOA curves for the TIP35C power transistor. I drew a line extending the SOA curve from the original 100V out to 200V following the deration. Out at 200V, with the current dropping by the square, the transistor could only handle 17mA! In theory, that is! Seeing as how that's less current that outputs are often biased at, it clearly wouldn't work. And this doesn't even take into account hot-spots on the die, as previously mentioned. Again, I understand these operational conditions. I have no intentions of pushing these devices anywhere near these levels. I do have 50 or so sets of TIP35C & TIP36C, but I still don't want to build an amp only to have it blow up, or worse, blow up my lovely speakers!
I just did a search, and it looks like MOSPEC actually makes TIP35 and TIP36 in D, E, and F parts too! Just like the 'C' part = 100Veo & 140Vcb, the D = 120Vceo & 160Vcb, E = 140Vceo & 180Vcb, and F = 160Vceo & 200Vcb. So the device OEM knows some of the parts can withstand a higher voltage. I just happened to stumble upon the specs with my NOS parts and a home-made breakdown tester. And, their deration curves look just like the one I drew and attached! Check out the attached datasheet to see the higher voltage part. Sure these aren't the best parts considering the wonderful options currently available, but since I have them, it's what I use to play around with.
I totally agree with Jan. Paralleling power output devices, and some other devices too, result in many improvements in an amp's performance. Some of the best amps in the audio world use many, many paralleled devices offering stunning performance. Granted there is added copper resistance, inductance, capacitance, etc., but all of that is totally cancelled out the devices' improved performance as each one is only handling only a fraction of the total work and through closed-loop feedback. Doug Self explains the reasons well in his book on power amp design.
Thank you for your input and dialog!
I've attached the SOA curves for the TIP35C power transistor. I drew a line extending the SOA curve from the original 100V out to 200V following the deration. Out at 200V, with the current dropping by the square, the transistor could only handle 17mA! In theory, that is! Seeing as how that's less current that outputs are often biased at, it clearly wouldn't work. And this doesn't even take into account hot-spots on the die, as previously mentioned. Again, I understand these operational conditions. I have no intentions of pushing these devices anywhere near these levels. I do have 50 or so sets of TIP35C & TIP36C, but I still don't want to build an amp only to have it blow up, or worse, blow up my lovely speakers!
I just did a search, and it looks like MOSPEC actually makes TIP35 and TIP36 in D, E, and F parts too! Just like the 'C' part = 100Veo & 140Vcb, the D = 120Vceo & 160Vcb, E = 140Vceo & 180Vcb, and F = 160Vceo & 200Vcb. So the device OEM knows some of the parts can withstand a higher voltage. I just happened to stumble upon the specs with my NOS parts and a home-made breakdown tester. And, their deration curves look just like the one I drew and attached! Check out the attached datasheet to see the higher voltage part. Sure these aren't the best parts considering the wonderful options currently available, but since I have them, it's what I use to play around with.
I totally agree with Jan. Paralleling power output devices, and some other devices too, result in many improvements in an amp's performance. Some of the best amps in the audio world use many, many paralleled devices offering stunning performance. Granted there is added copper resistance, inductance, capacitance, etc., but all of that is totally cancelled out the devices' improved performance as each one is only handling only a fraction of the total work and through closed-loop feedback. Doug Self explains the reasons well in his book on power amp design.
Thank you for your input and dialog!
Attachments
Measured some more devices
I was just salvaging some power devices from a damaged Sony amp and measured the Vceo and Vcbo ratings. The outputs used in the amp (see attached) were MN2488 and MP1620 Darlington devices.
MN2488 (NPN):
Rated: Vceo = 150V, Vcbo = 160V
Measured: Vceo = 205 to 220V, Vcbo = >320 (limit of my tester)
MP1620 (PNP):
Rated: Vceo = 150V, Vcbo = 160V
Measured: Vceo = 190 to 210V, Vcbo = 250 to 280
Again, with the SOA folding back so severely, these parts wouldn't be much good near their measured voltages. But, they would probably work well near, or even at, their rated voltages still having enough SOA (with enough devices paralleled) to work quite well.
What do you think?
I was just salvaging some power devices from a damaged Sony amp and measured the Vceo and Vcbo ratings. The outputs used in the amp (see attached) were MN2488 and MP1620 Darlington devices.
MN2488 (NPN):
Rated: Vceo = 150V, Vcbo = 160V
Measured: Vceo = 205 to 220V, Vcbo = >320 (limit of my tester)
MP1620 (PNP):
Rated: Vceo = 150V, Vcbo = 160V
Measured: Vceo = 190 to 210V, Vcbo = 250 to 280
Again, with the SOA folding back so severely, these parts wouldn't be much good near their measured voltages. But, they would probably work well near, or even at, their rated voltages still having enough SOA (with enough devices paralleled) to work quite well.
What do you think?
Attachments
seeing that datasheet ratings are lower than actual gives one more confidence...but i will still respect datasheet ratings...
"So the device OEM knows some of the parts can withstand a higher voltage. "
But the forward-biased safe area of operation still starts to fold back above 30V or so.
Unless you build a tester for forward-biased safe area of operation you know not what the part can do.
On Semiconductor 100% hand selects their TO-3 parts to meet the published specs. If it can't make the higher spec it is labled with a part number of a lower spec device.
TIP35/36 parts make a good choice for the top rails on an amplifier with tiered supplies, they can really push the current. They also work well on ±35V in bridge mode for car stereos.
"BC 556 "
Is used in diff inputs up to ±60V, never saw one used for an output. C grade have the best Hfe for diff pair use, typically around 500 at 1mA or so. The 2N5401 would be used when you need a higher voltage part, as would the MPSA93.
If you need to run 10mA at ±35V or so, a 2N5401 can have higher gain than a BC556A.
But the forward-biased safe area of operation still starts to fold back above 30V or so.
Unless you build a tester for forward-biased safe area of operation you know not what the part can do.
On Semiconductor 100% hand selects their TO-3 parts to meet the published specs. If it can't make the higher spec it is labled with a part number of a lower spec device.
TIP35/36 parts make a good choice for the top rails on an amplifier with tiered supplies, they can really push the current. They also work well on ±35V in bridge mode for car stereos.
"BC 556 "
Is used in diff inputs up to ±60V, never saw one used for an output. C grade have the best Hfe for diff pair use, typically around 500 at 1mA or so. The 2N5401 would be used when you need a higher voltage part, as would the MPSA93.
If you need to run 10mA at ±35V or so, a 2N5401 can have higher gain than a BC556A.
- Status
- Not open for further replies.