Yes! My mistake! The schematics are ok to make the pcb but this sim was done wrong. I corrected this in my last post. And I also updated the eagle schematics for both sides with the latest.
Changing the BC547s for BC639s in the bias stack definitely is better. And now that I also changed this in eagle, the even route better with less crossings.
I changed the Bs into Cs for the diff amps and T5/6, although I don't have the models for the higher hfe in the sims, using those in the actual build should be better, isn't it?
I also figured T8/9 might be better with higher Vce0 types, and T7/10 really don't need much there, as they never see their Vce going very high anyway.
Would using lower noise types be better on the current sources? The BC550/560 have too low Vce0 for the diff amps...
Here it is, with some tweaking done on R20 so the tails are much closer to each other. It's very difficult to go further, as now we're dealing with 100s of an ohms at a time and it jumps fast.
I think I got them close enough, since after tweaking R14/15 some more, they got even closer and the output offset is now below 4uV. I think going further is futile.
This closer symmetry allows looking closer at those much larger differences on the drivers, but after looking at this closer, I saw that we're looking at too different current paths, so I added a few more currents to compare better. The first difference, which isn't huge, but that will impact the difference on those drivers, is the currents in R25/28. Roughly 30uA of difference, times some beta, can make some difference on those drivers. Which may explain at least some, if not most of the difference in their emitter currents. It's tight for space there, but the 2 currents to compare would be their collectors.
The thing is, we have both drivers seeing that voltage drop on R25/28, but we must have a slight difference because on T16 we have that R30 collecting both emitter and R28's currents (and the output's bases), as opposed to R25 on top going straight to rail. This difference I suppose we can't avoid because we are quasi complementary and there is no other way.
I'm sure if I changed the - rails trannies to MJ2955s and re-wired for complementary topo, then we might have a much closer match.
This could be an other version of the amp. With only slight difference.
It's rather close now anyway. Look at the bias currents in those outputs. Not so bad.
I think I got them close enough, since after tweaking R14/15 some more, they got even closer and the output offset is now below 4uV. I think going further is futile.
This closer symmetry allows looking closer at those much larger differences on the drivers, but after looking at this closer, I saw that we're looking at too different current paths, so I added a few more currents to compare better. The first difference, which isn't huge, but that will impact the difference on those drivers, is the currents in R25/28. Roughly 30uA of difference, times some beta, can make some difference on those drivers. Which may explain at least some, if not most of the difference in their emitter currents. It's tight for space there, but the 2 currents to compare would be their collectors.
The thing is, we have both drivers seeing that voltage drop on R25/28, but we must have a slight difference because on T16 we have that R30 collecting both emitter and R28's currents (and the output's bases), as opposed to R25 on top going straight to rail. This difference I suppose we can't avoid because we are quasi complementary and there is no other way.
I'm sure if I changed the - rails trannies to MJ2955s and re-wired for complementary topo, then we might have a much closer match.
This could be an other version of the amp. With only slight difference.
It's rather close now anyway. Look at the bias currents in those outputs. Not so bad.
Attachments
This would make for a simulation closer to the real world. However what would interest me right now is a single frequency sim and take a look at the crossover point. I've been experimenting but haven't figured out yet how to get the single frequency instead of a sweep.
But an other sim shows a freq response that would be something like around 6hz to more than 100khz at -3db, if this is correct. That wouldn't be too bad.
I haven't looked at distortion yet, and the noise analysis didn't speak to me, I don't know how to interpret it and what to look for exactly.
I was just experimenting with a transient sim, but all I got was a good looking sine wave with no clipping at 30V peak. So I must not be doing things right.
I don't know the commands, I'm discovering them now.
"Oh well! It's an interesting one, because it's a bridge to begin with and it has a low number of parts. I wonder how that one would perform when there is a fault on the outputs, if the speakers would be at risk.
One also interesting feature is the fact that it's also a quasi. We could use 3055s throughout as well."
Back in its day the 3055 was king. I'm sure this would respond well to modern 3055's as well. I would probably use the Crown triple output circuit.
With respect to a fault on the output, I would add some sort of Triac crowbar circuit to protect the speakers. The Quad crowbar circuit would work fine. Peavey, QSC, etc, all have their versions.
Dawson Hadley would have been an interesting person to work for. Bongiorno made several bridge amplifiers (Ampzilla II, various Sumo, probably the Ampzilla 2000).
One also interesting feature is the fact that it's also a quasi. We could use 3055s throughout as well."
Back in its day the 3055 was king. I'm sure this would respond well to modern 3055's as well. I would probably use the Crown triple output circuit.
With respect to a fault on the output, I would add some sort of Triac crowbar circuit to protect the speakers. The Quad crowbar circuit would work fine. Peavey, QSC, etc, all have their versions.
Dawson Hadley would have been an interesting person to work for. Bongiorno made several bridge amplifiers (Ampzilla II, various Sumo, probably the Ampzilla 2000).
We're working on protection right now, but I don't know what would happen in case of an output failure, probably the same as any other push/pull.
But this is more than a simple bridge, it's a grounded one, and I don't know of any 3055s having been used in this way, so we're probably pioneering that. This makes it quite interesting to me.
Plus to see what we'll be able to get out of those old 3055s will be great. I think we can make quite a decent little amp, considering the use of old junk, and when we can get a bunch of watts out of it, with only 35V rails, this was not common back in the glory days of the 3055s.
And although 35V is kind of at or a little beyond the 3055's limit, at least we're not throwing it +-50V and calling it a 150W/8ohms amp.
I think I will attempt to verify the breakdown voltages of those 3055s I have on hand for that amp, to make sure 70V or close to it, won't be a problem. It would be a pitty to blow it right at the start.
That's exactly what we're doing, sticking to the plan, not changing to complementary, which could be done later in a slightly different version.
Sure, and if one can get a hold of those 3055H 100V types, then crank up the rails a little higher. (may have to use other diffs and stuff)
I don't like the crow bar concept myself, I think it's really radical, unless we make sure the outputs can withstand the short long enough to force a slow blow fuse to go out on the psu... This may actually be the case here, which is one of the things I would like to verify, because we're planning a 4 ohms load on the bridge, so the goal is 2ohms on those half amps and I'm hoping having 3 pairs can allow this.
I was trying to plot load lines on soa graphs just to do that, but I didn't get any feedback on my attempts yet, so I don't know if what I did was right. If it is, then we are well within the soa on 2ohms and we may have beefy enough output stages that can not only handle the 2ohms abuse, but perhaps also some longer term short abuses, which could maybe permit the crowbar approach.
I would prefer some output relay with some sensors...
I saw that one you posted earlier, and I would like to see a clearer schematic. It looks quite interesting for its low part number. It's not a flying rail type, but for its day, that must've been quite a thing.
For now, I'm trying to get a handle on some simulations, then we'll see about protection calculations, then I will want to revisit that load line vs soa thing, and actually this might be done while we do that protection calculation, since that's what it's supposed to be based on.
I was thinking about one more thing.
When we get this one working right, then we could look for a good schematic base to make a completely cascoded 3055 amp, still as a grounded bridge.
One of those could see its rails go up quite a bit and we could make sure we stay well within specs on the original 3055's Vce0.
We don't need to make use of modern and expensive other trannies in the other stages, there are old standards available that can do just fine. And without looking into exotic types we can't find.
This needs to be an easy enough build for diyers. And we can do this with a good pcb, making it simple and with as little wiring as possible.
When we get this one working right, then we could look for a good schematic base to make a completely cascoded 3055 amp, still as a grounded bridge.
One of those could see its rails go up quite a bit and we could make sure we stay well within specs on the original 3055's Vce0.
We don't need to make use of modern and expensive other trannies in the other stages, there are old standards available that can do just fine. And without looking into exotic types we can't find.
This needs to be an easy enough build for diyers. And we can do this with a good pcb, making it simple and with as little wiring as possible.
Crown did the grounded bridge circa 1974, and the first one used selected 3055's.
The MJ15015 is based on the 3055, the MJ15016 is based on the 2955. The first Motorola data sheets had all those part numbers together on the same data sheet.
Crown only used one set of four outputs on the first 200W/8Ω version. The Beta drooped so much that the driver transistors tended to blow up when driving 4Ω. Later they doubled the number of outputs and it became the Macrotech 600.
The crowbar idea works to protect the speakers, it assumes the amplifier is blown. If you use slow-blow fuses the Triac will blow too, as well as the emitter resistors for the outputs, but still protect the loudspeaker.
The Peavey CS800 is a good example of this in action. The customer brings one in for repair and askes for an estimate. You check the line fuse. If it is the correct one you tell him it won't be that bad. If the wrong fuse is in there you tell him it is going to be expensive. Later versions of this amp had a speaker relay in addition to foldback current limiting and a crowbar.
The MJ15015 is based on the 3055, the MJ15016 is based on the 2955. The first Motorola data sheets had all those part numbers together on the same data sheet.
Crown only used one set of four outputs on the first 200W/8Ω version. The Beta drooped so much that the driver transistors tended to blow up when driving 4Ω. Later they doubled the number of outputs and it became the Macrotech 600.
The crowbar idea works to protect the speakers, it assumes the amplifier is blown. If you use slow-blow fuses the Triac will blow too, as well as the emitter resistors for the outputs, but still protect the loudspeaker.
The Peavey CS800 is a good example of this in action. The customer brings one in for repair and askes for an estimate. You check the line fuse. If it is the correct one you tell him it won't be that bad. If the wrong fuse is in there you tell him it is going to be expensive. Later versions of this amp had a speaker relay in addition to foldback current limiting and a crowbar.
Last edited:
Depends on what caused the "DC to the speaker". If it's a fault in an input stage the SOA protection will still work, and it will get hot and shut down (if you have some sort of thermal cutout on the heat sink). If the fault is in the output stage it's already blown, so shorting everything and blowing the fuse won't do any more damage. I personally use relays on high power amps to open up in the event of DC to the speaker. You have to use a big one, but they show up on the surplus market for $5 all the time. The same realy can be used for power up muting, and another one for the soft start if you use 2kVA trafos.
Get the rails high enough, and 3055's won't be enough anymore. Not because of SOA or VCEO, but because of hFE (or the lack thereof). Maybe with 5, 6 or more in parallel per bank, but then you have to look at how many you have. And that's a lot of drilling and machining. I'm sure you could do it, but I'd be looking at using other output types. If someone had a large stash of lowish voltage high gain high current types, like 2N5630 or 2N6262 or even the MJ802 they'd be ideal. Such trannies would otherwise never find themselves in anything this powerful. Or go to cheap flatpacks like the TIP35/36, which have useable hFE at 20A. Those would be good for ease of assembly, and are still available anywhere.
Very cool! I didn't know that, so perhaps they were the only ones having done a grounded bridge with 3055s then, and although we're not going to be the only ones having done that, we'll be following in their footsteps.
Yes, that I knew, because I have that datasheet showing them together with basically the same specs but the MJs are 120V/180W and they are faster with a few other slight improvements. So I suppose we can consider the MJs as the glorified modern 3055 then.
This is nice, because we can swap the 3055s in our current design and just crank up the rails.
This is really a last resort and desperate feature, and if triggered even when the outputs aren't blown, then there is a good chance they will once this is triggered.
This is why I'm wary of this. I favor the output relay, although the presence of the output relay contact in series with the speaker does degrade the damping factor, some, the presence of the crowbar can also bring some disturbances, and when it activates and the outputs weren't dead, they surely have a good chance of becoming dead after that. I feel it's rather risky.
Now that sounds like something better. If the crowbar is used, we also need to add something to salvage whatever else we can and not let the crowbar blow everything.
I'm just thinking of the false positive triggers, which can happen, and then a perfectly still good amp gets turned into junk just because of that desperate measure.
We must protect expensive speakers for sure. There must be compromises.
Perhaps if the protection circuit is well calculated, plus maybe add a muting on the amp's input, or a big limiter at least, the crowbar can be safely activated without risking anything.
A wise precaution could also be the DC detection on the amp's output and have an output relay not come on if it's present. This would only protect when powering up but not after the amp has been hooked to the speakers by the relay.
And there is also one other thing that can be done as well, possibly not at a great cost, which is to use the other side of the SPDT relay contact to sense a short from the speaker line, and then not close that relay to hook up the amp to that short. When the short is removed, then this is detected and after a small delay the output can be sent out...
All this may be a little much for this project though. We would have to see how much extra needs to be added and how much more pcb real estate is needed for it.
The cost would probably not be that much, at least for the parts, as all this can be done with discrete and cheap parts. The pcb being a little bigger does get a little more expensive too, this needs to be evaluated.
Perhaps it's worth making the amp "safe", for a little extra money. When we look at all those diy amps made out there, how many take that many precautions??
This can be added without hardly extra cost and without doing it on the pcb. It requires only a simple thermostat wired into the transformer's primary to cut everything off and crew it on the main heatsink. We don't need to include this in the pcb design but that's definitely something to think about including when building an amp.
I am using 2 relays in parallel on my ongoing leach amp design. But the output is rarely physically near the input, and for good reasons, so although the relays could be triggered by the same circuit output, I would use separate relays. Besides, a beefy relay for the output is probably not the best choice for a muting one on the input.
Now you're making me think about the surplus shop "skycraft" in orlando. I was going there all the time looking for stuff. They have lots of good stuff there, ali baba!!!. Too bad there aren't any such shops where I am right now. I loved going to those surplus shops.
Well I don't think we're going to get anywhere close to that much power with this project the way it is now, but if we do make a version using those modern MJ versions of that 3055, with the higher rails, then we might come to that level of power, especially if we do a cascode topo!
We have a lot of fun in store for us!!!
Quite true! And that's why I think those newer glorified 3055s, the MJ15015/16 with their better performances could come in handy.
I think a version with flat packs might be rather easy and quick to make, once we have a good working design with the 3055s, we can take it as is and only change the outputs' layout for those flat packs. Shouldn't be too hard.
And talking about those alternative oldies with beefy specs. I have some old stock of BUX20. Think about those! 50A/350W and Ft min is 8Mhz. The only limiting factor is the 125V Vce0, but what a beefy one!!!!
Here are some questions that I'm asking myself while reasoning on some aspects:
Thinking about the input diffs first, their Vce0 exposure. One end, the collector (of any one of the diff amp's tr) is around a couple of volts below the rail where it is near, depending on which side we're looking at. And this won't vary very much apparenty. And the other end, the emitter, can only vary around the ground level in plus or minus the peak level of the input signal.
Am I on track so far?
So basically a diff stage transistor will only see at the maximum roughly half of the rail to rail voltage.
That's in a "classic" setup with the ground nicely in the middle and not moving.
Now let's think about this grounded bridge situation. The signal ground will follow the main ground regardless, so we can basically assume for the reasoning that it's at the same level. Now when the ground is being pulled around by the low side towards each rail back and forth, that ground level moving changes what Vce the diff stages can be exposed to.
This makes me think that since the ground gets pushed close to each rail back and forth, then the diff stages' transistors are seeing nearly the whold shebang and not just half.
Now tell me where my reasoning goes wrong, but that doesn't jive with what crown explain about the advantages of the grounded bridge, in that it reduces what transistors see of the rails, so their needs for handling high Vce would be reduced. Where I am going wrong in that reasoning?
Thinking about the input diffs first, their Vce0 exposure. One end, the collector (of any one of the diff amp's tr) is around a couple of volts below the rail where it is near, depending on which side we're looking at. And this won't vary very much apparenty. And the other end, the emitter, can only vary around the ground level in plus or minus the peak level of the input signal.
Am I on track so far?
So basically a diff stage transistor will only see at the maximum roughly half of the rail to rail voltage.
That's in a "classic" setup with the ground nicely in the middle and not moving.
Now let's think about this grounded bridge situation. The signal ground will follow the main ground regardless, so we can basically assume for the reasoning that it's at the same level. Now when the ground is being pulled around by the low side towards each rail back and forth, that ground level moving changes what Vce the diff stages can be exposed to.
This makes me think that since the ground gets pushed close to each rail back and forth, then the diff stages' transistors are seeing nearly the whold shebang and not just half.
Now tell me where my reasoning goes wrong, but that doesn't jive with what crown explain about the advantages of the grounded bridge, in that it reduces what transistors see of the rails, so their needs for handling high Vce would be reduced. Where I am going wrong in that reasoning?
WOW! I don't have any SOA info in 2 different datasheets on that one. Do you have one other?
This is surprising.
It's rather fast, super heavy current and total power, plus the thermal junction to case resistance of only 0.5K/W, that's an easy one to cool off compared to many.
Too bad. I'd like to see that SOA data.
Maybe for a power supply it would excel.
It's rated at 350W SOA at 20V.
The 40V figure may be a misprint. The 2N6277 is about 40W at 40V, so the BUX20 might be 60W at 40V?
Switching transistors have poor SOA for linear use, that's just the way it was.
http://pdf1.alldatasheet.com/datasheet-pdf/view/82712/SEME-LAB/BUX20.html
The I S/b line.
The 40V figure may be a misprint. The 2N6277 is about 40W at 40V, so the BUX20 might be 60W at 40V?
Switching transistors have poor SOA for linear use, that's just the way it was.
http://pdf1.alldatasheet.com/datasheet-pdf/view/82712/SEME-LAB/BUX20.html
The I S/b line.
Last edited:
I am trying to put together a spreadsheet to easily calculate the protections.
I'm having trouble figuring out how to proceed with the soa info.
For now, there is one criteria at least that I'm looking at, which is it must not be acting on valid signals. So since the bridge is destined to be loaded with 4 ohms, I am thinking about the peak output current that can be encountered. In the case of the 4ohms load, each side sees 2ohms, so I figured if the rails don't sag too much and we account for some losses in the Vcesat and emitter and other resistors, I'm just taking a rough figure of 32V peak on the output, which probably can never be reached unless the psu doesn't sag at all. On 2ohms, we're at 16Amps total already. The resistors to sense that current on are 0.22ohms, so here we are with a 3.52V drop to use to trigger the protections, a threshold under which we don't want to activate.
I can calculate the R18/19 and R62/63 to use this threshold, but then what to make of the R43/44?
And finally, there is the matter of R65/87 as well.
If I get this right, the R43/44 cause a knee in the protection locus and R65/87 add a second one.
Those calculations need to be done right, no room for guessing.
I'm having trouble figuring out how to proceed with the soa info.
For now, there is one criteria at least that I'm looking at, which is it must not be acting on valid signals. So since the bridge is destined to be loaded with 4 ohms, I am thinking about the peak output current that can be encountered. In the case of the 4ohms load, each side sees 2ohms, so I figured if the rails don't sag too much and we account for some losses in the Vcesat and emitter and other resistors, I'm just taking a rough figure of 32V peak on the output, which probably can never be reached unless the psu doesn't sag at all. On 2ohms, we're at 16Amps total already. The resistors to sense that current on are 0.22ohms, so here we are with a 3.52V drop to use to trigger the protections, a threshold under which we don't want to activate.
I can calculate the R18/19 and R62/63 to use this threshold, but then what to make of the R43/44?
And finally, there is the matter of R65/87 as well.
If I get this right, the R43/44 cause a knee in the protection locus and R65/87 add a second one.
Those calculations need to be done right, no room for guessing.
The diff pairs do get to see more voltage in the grounded bridge setup. But small signal transistors have no problem handling normal power at high Vce. Power transistors have second breakdown limitations, so they basically can't handle squat at or anywhere near rated Vceo. The higher the fT, the worse this is.
The Crown Macrotechs do not use diff pairs. They use op amps in the input stage, so the topology avoids high voltages in the front ends. With a 60 or 70V supply, MPSA06/56 are useable, as are 2N5401/5551. Lower rated devices (like the BC's) could be selected based on actual Icbo.
We're planning on the BC546/556 right now, with their Vce0 at 65V, we're not far from the max, although their Vcb0 is 80V... Those have been used in the leach amp by many, with rails at 58-60V I think, and nobody reported any failures.
I just measured again that dual 24V secondary transformer that I'm reserving for this amp. The unloaded voltage was 26.3V, and it's only a 160VA, which won't allow the max power on 4ohms loaded bridge, but should get fairly close to max power on 8ohms though. Once hooked up the the rest of the psu, that unloaded voltage may already sag a little bit before being actually loaded by a functioning amp. But regardless, I guess my rails might be close to 37V or slightly less at idle.
I'm sure the rails will sag quite a bit with increasing load, especially with an underpowered transformer. And at full power I expect the rails to fall below 35V for sure.
The option of using the mpsa06/56 or 2n5401/5551 is quite feasible for me, since I do have a bunch of those in stock as well. I even have tons of the mpsa92 but not the 42 right now. Situation that I will correct later, because I plan to use those for a leach amp. I don't even have any bc546/556 at all. I do have loads of bc547, but only few 557 (I just checked, I have 178 bc547, but only 3 557 left). So obviously to build this amp I will have to use other types at the time.
I would set it for 7A into a short circuit and 25A peak at (theroretically) zero Vce. That would be 75W each device at zero crossing (inductive) and 85W max dissipation (half way up the rail resistive). With 32V rails (64V total for the bridge). 3055's at reasonable temps can handle that for 25ms pulses (20 Hz bass, half cycle).
To calculate R43/44, just use KCL with the Vout set equal to the rail. The voltage drop across R62 must be the difference between the voltage across the emitter resistor and the Vbe of the protection transistor. When the limit current at "zero" vce is reached of course. Calculate the current required to get that, and you get a value for R43.