A modern 3055 handles full power to 40 volts, so second breakdown should not be an issue at 30 volt rails. It IS the reason you can’t use them on +/-100V, even if you select them for vceo above 200 (which you can do). So why do 3055 amps often blow above +/-20? They’ve got no GAIN. Hfe is less than 10 at 10 amps, I’ve measure 7 at 7 amps. And it’s in spec. 4 ohm load at 30 volts draws 7 amps peak, requiring an amp off the driver. Designers have found that fragile low-Cob drivers yield best results, both in sound quality and that pesky stability issue. Often they don’t get much of a heat sink, and if they end up heating up more (or just more quickly) than the outputs the bias regulator on the output heat sink is worthless. These amps more often die from thermal runaway or blown drivers taking outputs with them than from actual overload of the output transistors. Back in those days they didn’t have MJE1503x, and truth be told you really need something stouter like TIP41’s - mounted on the same heat sink as the outputs. The 2 ohm capable 3055 amp I threw together used a TRIPLE to get current gain, and TIP41 (read: takes 6 amps without dying) drivers. Some of those old RCA TO-5’s or the D40C stuff didn’t stand a chance - especially with a 2 watt slip-on heat sink, and older 3055s that might have had a gain of 5 at high current and low Vce.
That would be true for original parts actually made back then, but the newer made ones since that have been much better and actually all can handle well over their originally stated 60V limit.
So basically the designs are meant to make use of the newer parts that can handle this.
There is also the overhead to take into account, so +-30V total for the rails, doesn't mean they will ever be exposed to that much, even when 100% saturated. With rails sagging under load, some losses in wiring and whatever, plus the serious drop across the emitter resistors, the unloaded total 60V from the rails never make it completely to the Vce. So, starting with unloaded rails at, or about +-35 for a total of 70V, really brings them close to that "advertised" limit, but not so much in reality, under full load conditions, especially for those doing less than 8ohms. And we do know the parts made more recently will definitely handle more than that 60V from datasheets.
It's always a good idea of course, just in case, to check all of them first, for their Vce0, and that actually CAN be done with readily available (and pretty cheap) meters. So I personally will test all of them, as well as attempt matching as best I can, before using them.
Using such parts is definitely an option for many, depending on where they are in the world. The pcb design would need to be different, but the values calculated for the regular 2N parts should still apply, as the TIP types are hardly any different, perhaps only in dissipation capabilities, mostly due to the different case type.
Well, there are loads of such designs out there, so plenty to choose from. Including one from elektor with the quasi topo, and I actually have such a design for a grounded bridge based originally on that elektor design. Except the original design had to be heavily altered to fit the needs. I'm still working on that design and very recently I figured out a way to make it work, I think. It was working fine with simulations of each "side" but wouldn't work once bridged. I'm working on making use of an error amp to pair the sides up and this has a chance to work. This would be a very similar design, in a way, to this current one, and it would be a quasi.
Good point!
Just to see how the models are, I just ran a quick hfe test on it. (attached)
The gain does droop down to about that much, but may be a bit optimistic in the model compared to the real thing you've experienced.
We're aiming for 4 pairs, so we shouldn't get overly close to such high Ic though. It's something to keep in mind, but we are making use of the beefy drivers here.
A thermal sensing, not just the bias spreader's sensing, would definitely be a good thing to have, to make this as bullet proof as possible.
Still, aiming for rails at about 70V total when idling, really doesn't go very far above the breakdown limit anyway, and they "should" be able to handle that.
The main issue is to heatsink them well, with a good bias spreader mounting and add extra thermal sensing to take preventive actions.
If the power hungry folks who want to make use of the bridge in a bridged (bridged bridge) configuration, with nominal 8ohm load, each amp side sees a 2ohms (nominal) load, so that will definitely push this thing to its limits, but then it's basically a 1kW amp... Has to be properly protected thermally.
Attachments
Great!
Going to test this out right away.
If need be, we'll be able to tweak a few details and get a quick calculation tool that can be used to get new values when we tweak a few things, so it can be tested again without wasting time recalculating...
Finding the right res values that can actually be purchased is a must, hoping not to get into the very expensive types territory...
Still using 1% tolerance parts shouldn't be overly expensive, with metal film res. They don't need to be huge. (power wise)
Ok, I entered the values for the limiter, and tweaked a bit the currents in the sources/zeners/leds, that were a bit on the low side.
The current ID1 (D3,4,13,14 in sim) supposed to get to over 4mA, so I increased a bit the headroom for the zeners with the current sources.
It's great to calculate the dissipations in the spreadsheet. That gives a bird's eye view and where we need to be concerned where higher dissipation can occur, and it helps choosing proper res sizes.
I ran this sim at 20khz, and dropped the crowbar on the output.
The vas seems somewhat ok, and it does have current limit, that is at some 30mA or so. Should be fairly safe from destruction there.
The drivers and outputs are not being saved by the limiter though. It's going outside the SOA by a pretty large margin. So the effect of the limiter isn't strong enough. Wonder why. Maybe cfp config for the limiters might increase their effectiveness?
I'll run more sims at different frequencies...
The current ID1 (D3,4,13,14 in sim) supposed to get to over 4mA, so I increased a bit the headroom for the zeners with the current sources.
It's great to calculate the dissipations in the spreadsheet. That gives a bird's eye view and where we need to be concerned where higher dissipation can occur, and it helps choosing proper res sizes.
I ran this sim at 20khz, and dropped the crowbar on the output.
The vas seems somewhat ok, and it does have current limit, that is at some 30mA or so. Should be fairly safe from destruction there.
The drivers and outputs are not being saved by the limiter though. It's going outside the SOA by a pretty large margin. So the effect of the limiter isn't strong enough. Wonder why. Maybe cfp config for the limiters might increase their effectiveness?
I'll run more sims at different frequencies...
Attachments
Well then, it does help to disable the time constant at higher frequencies, but not enough.
Not too bad at 1k and even better at 20k though.
I think I'll try out using a cfp config on the limiters. Just in case.
Not too bad at 1k and even better at 20k though.
I think I'll try out using a cfp config on the limiters. Just in case.
Attachments
I think using CFP for the limiters is the way to go. With a far more effective action.
I tweaked the vas limiter slightly, which was aiming at ~30mA limit, but just like for the VI limiter, the issue is we take 600mV into account as the Vbe threshold, but that's not really where it actually happens.
It's more like 660mV. So in fact, the vas limiter only barely starts acting at 30mA, and I think the main reason why the VI limiter is letting too much go out of the soa, is probably also because it's acting later than calculated, because of the Vbe taken into account in the calculations.
So. I think this is where spreadsheets are so useful, because we can plug in a different value and quickly get corrected values for the resistors.
Let's try using 660mV instead of 600 for the Vbe turn-on ( Vbe(Q2) ).
And while at it, I checked the forward drop in the 4148 (referred as D1 in the calculations), which was barely over 450mV, which is a bit less than the 600 taken into account in the calculations.
So, I'm hoping that using such slight adjustments, the calculated values for the resistors will change a bit. Actually moving the locus a little, more to what reality would be, and maybe, just maybe, we might get closer to keeping things within the soa.
Using CFP is a huge help for the limiters. The singletons aren't able to sink sufficient current off the vas, which makes it hard to clamp the base current from the drivers.
I was getting less than 4mA of sinked currents, which isn't enough taken away from the drivers, and they keep pushing the outputs with them. I guess this is the amp's feedback fighting to keep things going...
Adding CFP on the limiters, greatly lighten the load on the first tranny, and the second can then sink a whole lot more from the vas, which for the + side was just a bit over 30mA, which in turn is just about where the vas limiter starts to want to act.
So, tweaking the vas limiter a bit also makes it a little more likely for the VI limiter to effectively sink enough to force it to act more clearly.
The vas is biased at nearly 15mA, so we can't be dropping the vas limit too low and too close to that.
More tweaking required, but getting much closer to effectiveness now.
I tweaked the vas limiter slightly, which was aiming at ~30mA limit, but just like for the VI limiter, the issue is we take 600mV into account as the Vbe threshold, but that's not really where it actually happens.
It's more like 660mV. So in fact, the vas limiter only barely starts acting at 30mA, and I think the main reason why the VI limiter is letting too much go out of the soa, is probably also because it's acting later than calculated, because of the Vbe taken into account in the calculations.
So. I think this is where spreadsheets are so useful, because we can plug in a different value and quickly get corrected values for the resistors.
Let's try using 660mV instead of 600 for the Vbe turn-on ( Vbe(Q2) ).
And while at it, I checked the forward drop in the 4148 (referred as D1 in the calculations), which was barely over 450mV, which is a bit less than the 600 taken into account in the calculations.
So, I'm hoping that using such slight adjustments, the calculated values for the resistors will change a bit. Actually moving the locus a little, more to what reality would be, and maybe, just maybe, we might get closer to keeping things within the soa.
Using CFP is a huge help for the limiters. The singletons aren't able to sink sufficient current off the vas, which makes it hard to clamp the base current from the drivers.
I was getting less than 4mA of sinked currents, which isn't enough taken away from the drivers, and they keep pushing the outputs with them. I guess this is the amp's feedback fighting to keep things going...
Adding CFP on the limiters, greatly lighten the load on the first tranny, and the second can then sink a whole lot more from the vas, which for the + side was just a bit over 30mA, which in turn is just about where the vas limiter starts to want to act.
So, tweaking the vas limiter a bit also makes it a little more likely for the VI limiter to effectively sink enough to force it to act more clearly.
The vas is biased at nearly 15mA, so we can't be dropping the vas limit too low and too close to that.
More tweaking required, but getting much closer to effectiveness now.
Just to illustrate the issue about the 4148 forward drops. Here's a reading on the currents and drops for the + and - side (amp high side), for D3 and D4 in the sim, which correspond to the D1 in the calculations.
The currents are pretty much in line with calculations, that say 4.32mA. Compare with the voltage readings peaking in the 4-4.5mA area. Not far off at all.
But the voltage drops are definitely closer to about 450mV than 600.
Taking this into account in the spreadsheet might nudge things a tad closer to getting the locus we seek. And get closer to keep things within soa.
The currents are pretty much in line with calculations, that say 4.32mA. Compare with the voltage readings peaking in the 4-4.5mA area. Not far off at all.
But the voltage drops are definitely closer to about 450mV than 600.
Taking this into account in the spreadsheet might nudge things a tad closer to getting the locus we seek. And get closer to keep things within soa.
Attachments
Here's a quick view of the tweaks and changes to CFP for the limiters.
The vas limiters act a tad earlier and possibly a little "stronger".
The CFP limiters sink the better part of some 30mA now, as opposed to barely 4mA with the singletons.
The sim ran at 20hz, and although it's still going out of soa, it's much much closer to bringing things in it...
The vas limiters act a tad earlier and possibly a little "stronger".
The CFP limiters sink the better part of some 30mA now, as opposed to barely 4mA with the singletons.
The sim ran at 20hz, and although it's still going out of soa, it's much much closer to bringing things in it...
Attachments
Removing the crowbar from the output, we can see the actual state of the limiter makes not intrusive.
Sim ran at 20hz, 8ohms load and 200Wrms power output.
Well within the soa all around, and thd shows the limiter not causing any troubles. Not bad so far. Now needs to get effective enough.
Sim ran at 20hz, 8ohms load and 200Wrms power output.
Well within the soa all around, and thd shows the limiter not causing any troubles. Not bad so far. Now needs to get effective enough.
Attachments
Looking at some old datasheets, one from ST states the 3055 hfe for Vce=4V at Ic=10A to be min 5, and for 4A is min 20.
My sim test fixture shows hfe of 24 for Ic=4A, so not exactly per datasheet, but not too far off. Maybe this can be "tweaked" in the model to reflect a more pessimistic real world possibility.
Will attempt lowering BF a bit..
I ran the amp sim with a 4ohms load (resistive of course), so each side sees a 2ohms load and the Ic is only peaking at nearly 3.3A, so the beta at that level is still above 20, at something around 28 or so. Using 4 pairs, this amp should drive this without too much trouble, even on hard loads. Making the bridged bridge a definite possibility.
My sim test fixture shows hfe of 24 for Ic=4A, so not exactly per datasheet, but not too far off. Maybe this can be "tweaked" in the model to reflect a more pessimistic real world possibility.
Will attempt lowering BF a bit..
I ran the amp sim with a 4ohms load (resistive of course), so each side sees a 2ohms load and the Ic is only peaking at nearly 3.3A, so the beta at that level is still above 20, at something around 28 or so. Using 4 pairs, this amp should drive this without too much trouble, even on hard loads. Making the bridged bridge a definite possibility.
Tried tweaking the model's BF down to 60 instead of 70.
This lowers the beta a little to a bit less than 22 for Ic=4A. This makes it more in line with ST's datasheet on that parameter.
Doesn't change things too much though. In the amp sim, the Ic on the 4ohms load (2 per side) barely reaches 3.3A. Still very good for our 4pairs.
This lowers the beta a little to a bit less than 22 for Ic=4A. This makes it more in line with ST's datasheet on that parameter.
Doesn't change things too much though. In the amp sim, the Ic on the 4ohms load (2 per side) barely reaches 3.3A. Still very good for our 4pairs.
I was wondering, I assume the values in the purple-ish color are manually chosen. This has to be done for the rail voltage anyway, and Re is also something we choose. But the value for R3 has the same color, so does that mean that one is hand picked? Knowing the number of pairs, the actual value would be chosen and divided accordingly, to get that value into the calculations of others...
If this is the case, perhaps would could pick a value a bit higher, so R2 would also be a bit higher and those are the ones that can have the most influence on the circuit around when the limiter isn't activated. This is to maximize the transparence of the limiter..
The sensing speedup cap also comes into play at the higher frequencies, and I've been tweaking that and found a sweet spot for it. It was 10n and it's now 27n, which helps quite a bit with frequencies above some 2k or so.
I tried putting the time constant cap back, but it always wreaks havoc, no matter what its value, so that one is better kept off the limiter. I didn't put any time constant either on the vas limiter.
Since I've been running far more sims at many other frequencies, I find that it's only at some frequencies that there is the most trouble. The rest of it is fine.
By the way, to test the limiter, I put the amp in its worst scenario, with a hard short crowbar on the output, with signal to drive the amp at full power. So if we can tweak the limiter to keep it safe in such conditions, this thing should be fairly safe. And with an input signal limiter, triggered by the differences on the input stage ltp, this can make it even safer.
Last edited by a moderator:
spookydd
The values Vcc, Re, R3* and Vz are all chosen, as described by MK. R3* is multipled by N, the number of pairs of outputs; in your case 4, so each R3 is 910Ω. The spreadsheet is based on one output pair.
Those in red are the nearest standard values to the calculated values derived from the VI breakpoints spreadsheet that offer close conformity to the original breakpoints.
Brian
PS. I’ve posted the spreadsheets to you.
The values Vcc, Re, R3* and Vz are all chosen, as described by MK. R3* is multipled by N, the number of pairs of outputs; in your case 4, so each R3 is 910Ω. The spreadsheet is based on one output pair.
Those in red are the nearest standard values to the calculated values derived from the VI breakpoints spreadsheet that offer close conformity to the original breakpoints.
Brian
PS. I’ve posted the spreadsheets to you.
One more parameter could be added, the number of pairs used, and so instead of entering the sense res value divided by the number of pairs, this could be automated and the value entered for R3 would be the actual res value used.. Just an idea to make it one more step automated.
Got that, thanks. Super useful. A moment to figure out how it works and I'm tweaking some values that I'm about to test in sim now.
In the sim, I just put single zeners in the limiter, but it would be best to use strings of lower voltage ones, in the 5.4-6.4V range for lower thermal drift.
I don't have that many zener models for the sim, so I'm faking it by using ako with a tweak to the voltage. But in a build, the 12V ref could be better off at 12V4 if made of 2 6V2 zeners.
That's one of the tweaks I'm doing, and one other is for the 18V ref, which I tweaked to 17V6, made of 2 5V6 and a 6V4 zeners.
Also going to try out overall higher res values, for less disturbance (more transparence) to the amp, while at the same time lowering dissipations.
MK's analyses are based on one R3 only, so I would end up having to divide by N (where N is the number of pairs) in the spreadsheet calculations. Showing the values based on one pair, as stated in the cells E,F,9,10, to me, is less confusing.
So if the R3 value is not for one pair, then the calculations, as they stand, fail drastically. You won't get the locus you're after. Try it. I have in some sheets (not the ones posted) added a line that shows the value N*R3 , but the statement
"*For N protected pairs of o/p devices, R3 actual = R3xN"
should be obvious. Anyway, you can add that "calculation" yourself.
Brian
We have a working protection on this amp.
I ran a lot of sims, at various frequencies, loads, shorted output, etc...
I found a compromise that uses the time constant caps and the speedup ones on the limiter, with an added tiny cap to damped somewhat the funny business going one when the limiter is acting. This is inherent to the limiter to cause all kinds of oscillations, and they just can't be going away, however they can be limited to more reasonable levels.
I wonder if, when acting at high frequencies, like above 10khz, the small amount of activity going a bit above the soa limit (derated dc), is much of a concern. As this is just some top ends of fast transients, short duration, although repetitive on a sine drive signal, this wouldn't be the same on music signal...
I didn't put a time constant on the vas limiter, which acts a bit later anyway. I found that having both the vas limiter and vi limiter acting at the same time causes more havoc and they seem to interfere with each other.
The limiter on the vas is only to protect the vas, and there is sufficient headroom there. When the vi limiter is acting on a dead short, with full power sine wave drive signal applied, the vas never comes close to soa limits, and its limiter doesn't need to act.
I cleaned up the schematic sim, after having tried many things that messed it up visually. Hopefully it's clean enough now to be understandable.
Regarding the vi limiter, I raised the overall values for the resistors in the limiter, so it lowers quite a bit the dissipations while bringing less disturbance to the amp's circuit because of its presence, from all the bias currents needed.
I also changed the reference zeners to be strings of lower voltage ones instead of single zeners. I keep the zeners in the voltage range of 5v6 to 6v4, for best temperature coefficient.
Using Brian's spreadsheet to get the values instantly is a great tool. Thanks brian
I didn't have the proper zener models to simulate with the actual strings, so I used single zeners but with an altered voltage to match what the strings would produce.
With all the sims I ran to push it hard to the limits, and check how intrusive the vi limiter is on legit signals, I found the compromise to be acceptable. On the nominal load of 8ohms (resistive), it doesn't act at all and it pretty much transparent, not bringing in much, if at all, disturbance causing any additional distortion to the amp, all the way to full power (~200Wrms).
Testing it on a 4ohms load (resistive), there are of course more losses in the overhead, mostly in the emitter resistors and the rail sagging (I put a 200mohms series resistance on the supplies), which makes for a slightly lower full power level than the theoretical 2x. So we don't get quite twice the 200Wrms from 8ohms nominal, but rather about 370-375Wrms (4ohms nominal). And the vi limiter doesn't intervene here either, which is a planned legit signal.
This 4ohms nominal load on the amp would be like an 8ohms nominal on the bridged bridge configuration.
Now taking into account the impedance drooping from nominal and a reactive load, if we take 6ohms instead of 8ohms for nominal, that passes easily. And take 3ohms for the 4ohms nominal load, that comes to just about the limit where the vi limiter wants to start acting.
On anything lower than 3ohms, the vi limiter will start acting, but anything beyond there would be considered not to be legit usage, so whatever extra distortion the limiter causes can't be complained about.
I've done some tests on a possible input limiter as well, which watches the amp's ltp for imbalance and would compress the input drive signal to limit it. This can work, and I would really like to use a vactrol to apply the compression. Those vactrols are getting rare and difficult to find nowadays. Don't know why. It's possibly the lowest distortion method to compress an input signal, and due to its inherent time constants, it's a natural choice for that purpose.
I think for many it might be a good idea to make their own vactrols...
I ran a lot of sims, at various frequencies, loads, shorted output, etc...
I found a compromise that uses the time constant caps and the speedup ones on the limiter, with an added tiny cap to damped somewhat the funny business going one when the limiter is acting. This is inherent to the limiter to cause all kinds of oscillations, and they just can't be going away, however they can be limited to more reasonable levels.
I wonder if, when acting at high frequencies, like above 10khz, the small amount of activity going a bit above the soa limit (derated dc), is much of a concern. As this is just some top ends of fast transients, short duration, although repetitive on a sine drive signal, this wouldn't be the same on music signal...
I didn't put a time constant on the vas limiter, which acts a bit later anyway. I found that having both the vas limiter and vi limiter acting at the same time causes more havoc and they seem to interfere with each other.
The limiter on the vas is only to protect the vas, and there is sufficient headroom there. When the vi limiter is acting on a dead short, with full power sine wave drive signal applied, the vas never comes close to soa limits, and its limiter doesn't need to act.
I cleaned up the schematic sim, after having tried many things that messed it up visually. Hopefully it's clean enough now to be understandable.
Regarding the vi limiter, I raised the overall values for the resistors in the limiter, so it lowers quite a bit the dissipations while bringing less disturbance to the amp's circuit because of its presence, from all the bias currents needed.
I also changed the reference zeners to be strings of lower voltage ones instead of single zeners. I keep the zeners in the voltage range of 5v6 to 6v4, for best temperature coefficient.
Using Brian's spreadsheet to get the values instantly is a great tool. Thanks brian
I didn't have the proper zener models to simulate with the actual strings, so I used single zeners but with an altered voltage to match what the strings would produce.
With all the sims I ran to push it hard to the limits, and check how intrusive the vi limiter is on legit signals, I found the compromise to be acceptable. On the nominal load of 8ohms (resistive), it doesn't act at all and it pretty much transparent, not bringing in much, if at all, disturbance causing any additional distortion to the amp, all the way to full power (~200Wrms).
Testing it on a 4ohms load (resistive), there are of course more losses in the overhead, mostly in the emitter resistors and the rail sagging (I put a 200mohms series resistance on the supplies), which makes for a slightly lower full power level than the theoretical 2x. So we don't get quite twice the 200Wrms from 8ohms nominal, but rather about 370-375Wrms (4ohms nominal). And the vi limiter doesn't intervene here either, which is a planned legit signal.
This 4ohms nominal load on the amp would be like an 8ohms nominal on the bridged bridge configuration.
Now taking into account the impedance drooping from nominal and a reactive load, if we take 6ohms instead of 8ohms for nominal, that passes easily. And take 3ohms for the 4ohms nominal load, that comes to just about the limit where the vi limiter wants to start acting.
On anything lower than 3ohms, the vi limiter will start acting, but anything beyond there would be considered not to be legit usage, so whatever extra distortion the limiter causes can't be complained about.
I've done some tests on a possible input limiter as well, which watches the amp's ltp for imbalance and would compress the input drive signal to limit it. This can work, and I would really like to use a vactrol to apply the compression. Those vactrols are getting rare and difficult to find nowadays. Don't know why. It's possibly the lowest distortion method to compress an input signal, and due to its inherent time constants, it's a natural choice for that purpose.
I think for many it might be a good idea to make their own vactrols...
