Interesting. This will have to go on the list of rigs to lash up. ....and will have to acquire some parts to test.
In general, you just about never find this sort test happening in a production environment because there is no way to get the heat out of a wafer. When a wafer undergoes a high current test, it is pulsed to keep the average power over time way down low.
Where this is guaranteed, it is sample tested at the package level (some # of parts per lot) for 1 second duration. Not all parts have a guaranteed s/b spec - even though it is listed for the 3055 it isn’t tested. Same for high performance parts like the 4281. Only the MJ15024 and 2N3773 say “100% tested”. Even then it is probably only a sample - but usually well correlated to other parameters that are 100% tested (on wafer).
If they're onsemi parts, not much of a surprise for sure. But from other manufacturers, it's interesting...
That's an interesting way to approach it. For those without a meter that can do it, it can be rigged.
From what I understand about Vce0, it's supposed to be done with the base open. This is also done for the cb junction, and the emitter is left open.
What you're describing seems a little more involved and is definitely something to handle with great care to avoid destruction. A more real world test.
If we went strictly by the book (the datasheet), the Vce0 of 60V would impose a rail to rail voltage limited to about 60V, or perhaps a tiny bit more because there are losses and overhead preventing the whole spread to be applied to the ce junction. In the design in question here, I go for 70V rail to rail when idling at no power output, and knowing the rails will sag a little at full power and there are losses and overhead, the ce junctions aren't subjected to much more than 60V.
When at the onset of clipping, the sims have shown the ce junction to experience 65V of swing, which is 5V more than the datasheet rated Vce0, but that likely won't be an issue with the parts. We're really not pushing them very much beyond their rating and we know they do handle more than that, most of them...
For the sims, I put an impedance of 0.1 or 0.2ohms on the power supply, which probably is similar to a real supply. Or the real thing may even be better, with a beefy transfo, big rectifiers and huge filter caps with low ESR and all that...
All very interesting stuff. Very good to discuss this and thanks for the insights.
I suppose the test I'm doing with that meter may not be exactly what it should be, but doesn't reflect to some degree what voltage the parts can withstand?
If they pass this test with a voltage much higher than what they're rated for, doesn't that insure they will be fine with the voltages we subject them to?
All I want is to make certain that there won't be a problem with the applied rail voltages, when the amp is at max power and the vce gets 65V. That's only about 5V of losses and overhead. Maybe a real supply might sag a little more and so the actual max power on a real amp might be a little lower than the sims show.
In any case, the amp must be robust and not break at first solicitation.
The last major hurdle on this design is to get the protection working properly.
I tried a little something yesterday, I put an extra transistor on the input stage current source to bring down the tail current, thinking when the protection acts, it would cause a drop of tail current and prevent the fight agains the protection from acting.
But what I find is that the diff stage keeps amplifying pretty much all the way until its tail current is totally gone, so there is no way to smoothly control this and this seems like a bad idea to use for protection.
The main issue in this design when it comes to VI protection, is that although we sense only on the high side of the bridge, the action should be done on both sides, and that's not as easy as it should be.
The bridge has separate stages for output, drivers and vas, but only a single input stages.
That's one of the beautiful things about this architectures.
But how to apply an action from protections remains to be devised.
There is one other design that I played with, using 3055/2955, and it also works great in sims, which is based on the bryston architecture. It shows even greater performance, but not the same power output as a bridge. Still very interesting to see how much we can get out of those older type devices.
One thing that would be really great is being able to obtain a large number of devices and test them for their characteristics, to make sets of the most identical parts to make amps out of.
I know that's what bryston does for their own amps. They get huge lots of parts, and test them all to make sets. So their amps have really good symmetry and work nicely.
I suppose the test I'm doing with that meter may not be exactly what it should be, but doesn't reflect to some degree what voltage the parts can withstand?
If they pass this test with a voltage much higher than what they're rated for, doesn't that insure they will be fine with the voltages we subject them to?
All I want is to make certain that there won't be a problem with the applied rail voltages, when the amp is at max power and the vce gets 65V. That's only about 5V of losses and overhead. Maybe a real supply might sag a little more and so the actual max power on a real amp might be a little lower than the sims show.
In any case, the amp must be robust and not break at first solicitation.
The last major hurdle on this design is to get the protection working properly.
I tried a little something yesterday, I put an extra transistor on the input stage current source to bring down the tail current, thinking when the protection acts, it would cause a drop of tail current and prevent the fight agains the protection from acting.
But what I find is that the diff stage keeps amplifying pretty much all the way until its tail current is totally gone, so there is no way to smoothly control this and this seems like a bad idea to use for protection.
The main issue in this design when it comes to VI protection, is that although we sense only on the high side of the bridge, the action should be done on both sides, and that's not as easy as it should be.
The bridge has separate stages for output, drivers and vas, but only a single input stages.
That's one of the beautiful things about this architectures.
But how to apply an action from protections remains to be devised.
There is one other design that I played with, using 3055/2955, and it also works great in sims, which is based on the bryston architecture. It shows even greater performance, but not the same power output as a bridge. Still very interesting to see how much we can get out of those older type devices.
One thing that would be really great is being able to obtain a large number of devices and test them for their characteristics, to make sets of the most identical parts to make amps out of.
I know that's what bryston does for their own amps. They get huge lots of parts, and test them all to make sets. So their amps have really good symmetry and work nicely.
One other thing I was wondering about. Regarding the SOA and the second breakdown issues:
If we plot (in sims) the usage against the rated SOA (derated for heating), as long as it stays well within that SOA available and don't get close to the second breakdown limits, regardless of load, how complex it is and how hard it gets pushed when output is short circuited or amp is overdriven, like clipping hard, there should be no concern about second breakdown issues.
On complex loads, it can get close to some limits, but if we make sure it's not taken beyond those, it should work fine.
It's a tricky thing to put proper VI protection on this architecture and make it work as intended, but once this is made to work, that amp design should be good, robust and reliable.
I've never seen any 3055/2955 amp designs giving true (rms) 200W output (on 8ohms). Without totally trashing the signal along the way...
This design would do this, and have quite decent performances.
If we plot (in sims) the usage against the rated SOA (derated for heating), as long as it stays well within that SOA available and don't get close to the second breakdown limits, regardless of load, how complex it is and how hard it gets pushed when output is short circuited or amp is overdriven, like clipping hard, there should be no concern about second breakdown issues.
On complex loads, it can get close to some limits, but if we make sure it's not taken beyond those, it should work fine.
It's a tricky thing to put proper VI protection on this architecture and make it work as intended, but once this is made to work, that amp design should be good, robust and reliable.
I've never seen any 3055/2955 amp designs giving true (rms) 200W output (on 8ohms). Without totally trashing the signal along the way...
This design would do this, and have quite decent performances.
I found I had to be careful with buying new 2n3055's as they tended to be a better bandwidth and better Hfe than originals from the 70's.
I bought a Maplin 225WRMS 2n3055/mj2955 amp without output transistors.
So I bought new ones and the thing oscillated nicely.
I increased the VAS capacitor a little and that calmed things down.
I bought a Maplin 225WRMS 2n3055/mj2955 amp without output transistors.
So I bought new ones and the thing oscillated nicely.
I increased the VAS capacitor a little and that calmed things down.
Just out of interest I tested the breakdown voltage of a modern 2n3055 and it was 115 volts.
Yes, those newer parts are faster and better overall, and older designs specifically made for the older ones weren't made for the faster devices.
But we're designing new amps now with the newer devices in mind, so hopefully oscillations are handled and we won't have any issues.
It's good that they're faster, so we get a better bandwidth and the high end is better than it would be with those older parts. Our sim models must be taking into account the faster specs, with a higher Ft and all that. So I hope our sims are accurate as much as possible.
The models aren't perfect, especially if they're modpex, but we've done some tweaking, at least for a few parameters, and I hope we're closer to reality there.
That's not bad. Pretty much in line with what I found myself on the majority on my 3055s and the 2955s as well. Hardly any had less than 120V, most around 125-127 or so..
This may not be the full reflection of the actual second breakdown though, but I hope it means they will withstand more than 60V Vce.
And since we're not aiming at very much over the rated datasheet specs, we should be in the "safe" zone when it comes to Vce voltage.
If someone who wants to build this thing and they're wary of breaking the limit, then it's fairly easy to size up the psu for 60V instead of 70, but I don't think this will really be necessary.
200W/8ohms from a 3055 based amp, with decent performance, this I'd like to see coming real.
I've been thinking it might be good to try to add a limiter to it, to deal with overload and prevent hard clipping, to protect speakers.
And a thermal sensing might also be a good idea, so this beast can be made super reliable.
It can be made into a very nice build for amateurs amp builders, with simplicity, with as few wires as possible, all on a single pcb.
I don't think we can do it with one a single heatsink though, not only because it's hard to find the right kind that big, and it's a lot of devices to put on a single heatsink. And the heatsink needs to be beefy enough with a low enough thermal resistance. Personally I'd go for that single flanged one from conrad heatsinks, which can be 30cm long, and use 2 of them. I did some cad on this before, and 8 devices can fit on there, with the drivers...
After thinking about your test DIY test rig, I am wondering what the second breakdown failure mode looks like. The part is cooking at 2.87A with 12 to 15v on the base with an appropriate resistor from emitter to ground and 40V Vce the collector, and then what? Does collector current spike at second breakdown?
Another question, if you are using 15v at the base, then you will have nearly 15v at the emitter. If you want 40v Vce, then you would need 55V from collector to ground to get 40v Vce, yes?
It would be helpful to have a written series of steps and a sketch of the test circuit.
I will be using a pair of variable power supplies in series that are each 35v and can source up to 20A. They have an adjustable current limit.
I was wondering if anyone had looked into using supercaps for an amp psu filtering, instead of the regular electrolytic caps, or perhaps in a combination...
They are much smaller for the same capacity, although they have overly low voltage capability, they can be put in series to compensate for that, and they could be providing huge reserve. But how do they perform in such a usage? That is the question.
They are much smaller for the same capacity, although they have overly low voltage capability, they can be put in series to compensate for that, and they could be providing huge reserve. But how do they perform in such a usage? That is the question.
The super caps I have seen are low voltage. I believe that they are intended for low current draw for a long time at relatively low voltage.
Actually they're not aimed specifically at low currents. They can be discharged and recharged very fast, just like any caps are. But their main drawbacks are their very low voltage.
However, since they're very small for their capacitance, they can be arranged in series to increase the voltage, lowering the resulting capacitance in the process, but since they have such very large capacitance, it would still be quite a bit left once in series.
Quite a few are needed to get to a voltage like 75-80V though...
The thing I'm wondering about is how they behave as filtering caps, and things like their ESR and whatever...
They're definitely not made for AC usage, but filtering?
However, since they're very small for their capacitance, they can be arranged in series to increase the voltage, lowering the resulting capacitance in the process, but since they have such very large capacitance, it would still be quite a bit left once in series.
Quite a few are needed to get to a voltage like 75-80V though...
The thing I'm wondering about is how they behave as filtering caps, and things like their ESR and whatever...
They're definitely not made for AC usage, but filtering?
Anyone knows of a good source for regular filtering caps with some 75-80V that won't cost a fortune?
I've been doing sims thinking of some 88000uF worth of filtering, and as it's better to use a larger number of smaller caps instead of fewer big ones, I've been trying to compare. But the costs are quite high when getting into such high voltages and the available values are fewer as well.
I looked at arrow, mouser, future elec, vishay, sparkfun (they don't have much of that) and have yet to find a good choice for a reasonable cost.
I've been doing sims thinking of some 88000uF worth of filtering, and as it's better to use a larger number of smaller caps instead of fewer big ones, I've been trying to compare. But the costs are quite high when getting into such high voltages and the available values are fewer as well.
I looked at arrow, mouser, future elec, vishay, sparkfun (they don't have much of that) and have yet to find a good choice for a reasonable cost.
Second breakdown itself is actually non destructive. You get localized hot spots which reduce vbe very quickly. Then the current runs away, which kills the transistor from overheating. A real S/B tester detects the drop in vbe and shuts the collector current down before it has time to die. You keep turning it up the stress till it actually does it.
My makeshift rig would be destructive if the transistor doesn’t pass. So what? If it doesn’t pass spec, then you don’t use any out of that batch. I will test some parts beyond spec, to see what they will actually do - for example to see if a batch of surplus 2N5631’s will pass the 1.5A/100V 2N3773 torture test. Most of the time they will - crack one of each open some time and you’lll see why. But most of the time it’s just to test a handful of parts to see if they are real or fake, when I can’t afford to waste a good one. If they’re good they pass and you can still use them. If not you haven’t really lost anything.
If you’re testing for s/b at power levels significantly below rated power, say at a relatively high voltage at low current, you can rig fold back current limiting into the collector supply and let that limit things when it does run away. If thermally limited dissipation is respected it will survive - just jump up in current sharply till the supply limits.
You can use 3055’s up to 80 volts, if you verify VCEO. If it handles full power up to the normal single rail voltage ( in this case 40 volts), and you’ve sized them appropriately for dissipation, you will not run into second breakdown issues with loudspeaker loads at audio frequencies (20 Hz and up).
I only buy soup can caps surplus, period. They just want too much for them otherwise. Buy a couple when you find em, see if they form correctly, then get more. Form them when you get them, keep a stockpile, and store them inside in the AC after a complete discharge. Properly stored, shelf life is long. Can’t say as much for the typical little 10 uf leaded electros. Those do go bad rather quickly.
I had a look at a 2N3055 datasheet and it specifically states S/B at 40V at less than 3A for 1 second. From your explanation, I am guessing that a collapse of Vbe during that 1 second would indicate a S/B failure, yes?
I searched Digikey and Mouser. The prices for 2N3055 are $5-$6 each. That is 2X the price of a good power MOSFET. Is there a cheap source of good 2N3055? Jameco has some cheap 2N3055 but I am not sure of the quality.
2.87 A, which is 115 watts at 40 volts. 1 second is about as long as you can put in 115 watts without a heat sink, before case temp skyrockets. If you could keep the case temp at 25, it would do it all day. Putting it on a heat sink with the fins soaking in an ice bath would do it. Pulsed SOA is higher because of the thermal time constants involved. Audio waveforms can make use of a 10 millisecond curve safely, provided TJ is also respected. For conservative operation of subwoofers, use the 100 ms if one exists.
The collapse of vbe is the telltale sign that the hot spot is forming. This does take time, depending on how localized it is. The pulsed curves are much higher on units that are less susceptible to 2nd breakdown - heating is more uniform and takes longer to heat up and run away. 100 ms curve on Sanken LAPTs is more than double the DC curve, and nowhere near that on the 3055 - even the 10 ms curve isn’t at 2x DC.
Distributors list prices seem to have gone through the roof on all TO-3’s lately. It may be a sign they are all nearing EOL. I’ve bought a lot of transistors over the last 3 years just to have lifetime supplies. Been buying up lots of surplus, too. When the 2N5879/5881 pairs went to 1.49 at BG Micro I jumped on them and got 40 pair. In 2 orders, the first was to dertermine if they were real. Aluminum TO3’s always are, but you’ve got to check. Only rated 60V, but I’ll bet you can run em higher. S/B point is at 40 volts, same as 3055. And typical beta is 15 at 15 amps vs. only 7 for the 3055. If I end up building one of these grounded bridges, you know what I’ll be using for outputs. The other application I had in mind was rail modulators for a nice 100 watt class H i prototyped a few years back. I’ve got a dozen of the transformers (each amp takes two) and want to make a batch some day.
The collapse of vbe is the telltale sign that the hot spot is forming. This does take time, depending on how localized it is. The pulsed curves are much higher on units that are less susceptible to 2nd breakdown - heating is more uniform and takes longer to heat up and run away. 100 ms curve on Sanken LAPTs is more than double the DC curve, and nowhere near that on the 3055 - even the 10 ms curve isn’t at 2x DC.
Distributors list prices seem to have gone through the roof on all TO-3’s lately. It may be a sign they are all nearing EOL. I’ve bought a lot of transistors over the last 3 years just to have lifetime supplies. Been buying up lots of surplus, too. When the 2N5879/5881 pairs went to 1.49 at BG Micro I jumped on them and got 40 pair. In 2 orders, the first was to dertermine if they were real. Aluminum TO3’s always are, but you’ve got to check. Only rated 60V, but I’ll bet you can run em higher. S/B point is at 40 volts, same as 3055. And typical beta is 15 at 15 amps vs. only 7 for the 3055. If I end up building one of these grounded bridges, you know what I’ll be using for outputs. The other application I had in mind was rail modulators for a nice 100 watt class H i prototyped a few years back. I’ve got a dozen of the transformers (each amp takes two) and want to make a batch some day.