Here is the resistor choices from the spreadsheet.
With the zener strings, all made from the 5v6 to 6v4 range only.
All the limiter resistors, except for R1, can be regular 1% or better, metal film. And R1 might be best as a 1W, to be on the safe side.
One thing that I did do, is to choose a rail voltage a tiny bit over the 35V, to be on the safe side, so the locus crosses the Vce axis beyond the rail voltage (as per M.K's recommendation)
The time constant (C1/R8) is at 220us, and the speedup cap has the same 22n value. With the small 15p cap on the CFP limiter, all this keeps the funny business down to a reasonable level when the limiter is acting.
By the way, the choice of LED based current sources for the limiter references is for thermal drift reasons.
With the zener strings, all made from the 5v6 to 6v4 range only.
All the limiter resistors, except for R1, can be regular 1% or better, metal film. And R1 might be best as a 1W, to be on the safe side.
One thing that I did do, is to choose a rail voltage a tiny bit over the 35V, to be on the safe side, so the locus crosses the Vce axis beyond the rail voltage (as per M.K's recommendation)
The time constant (C1/R8) is at 220us, and the speedup cap has the same 22n value. With the small 15p cap on the CFP limiter, all this keeps the funny business down to a reasonable level when the limiter is acting.
By the way, the choice of LED based current sources for the limiter references is for thermal drift reasons.
Attachments
I've been trying to find the best way to make an input limiter for the grounded bridge, and tested a few things, and this would work nicely in concert with the vi limiter, that seems so far impossible to keep things 100% within the SOA. Having compression on the input drive signal would take care of the rest, and so the vi limiter working alongside the input limiter would make it possible.
The input limiter can double up as input muting, and it would work in other situations where the vi limiter doesn't really enter into action, such as when the amp starts clipping.
So the input limiter would be a great feature to have.
The thing is, I've been trying to find alternative to the vactrols because those are getting so difficult to obtain, and I don't think there is anything that's good enough, while staying simple.
The complex ICs like those from dbx or others (that corp, etc...) are not only too complex, they're expensive and not necessarily obtainable in some places on this planet. Plus I don't find them to be that good anyway, compared to the much simple and cheaper vactrol option.
I thought about the FETs, although they're not so linear, but clamping a symetric signal without referencing to ground isn't that easy to do with those, and the ordinary JFETs don't have an Rdson low enough anyway.
Using opamps, I didn't find anything worthwhile.
Using BJTs could work, keeping things mostly discrete, and they do cause distortion, not sure how much in comparison with FETs, but I've been exploring this and it seems overly complex and not so easy to keep linear.
So I don't see any other viable solution besides trying the vactrols anyway...
Any ideas?
The input limiter can double up as input muting, and it would work in other situations where the vi limiter doesn't really enter into action, such as when the amp starts clipping.
So the input limiter would be a great feature to have.
The thing is, I've been trying to find alternative to the vactrols because those are getting so difficult to obtain, and I don't think there is anything that's good enough, while staying simple.
The complex ICs like those from dbx or others (that corp, etc...) are not only too complex, they're expensive and not necessarily obtainable in some places on this planet. Plus I don't find them to be that good anyway, compared to the much simple and cheaper vactrol option.
I thought about the FETs, although they're not so linear, but clamping a symetric signal without referencing to ground isn't that easy to do with those, and the ordinary JFETs don't have an Rdson low enough anyway.
Using opamps, I didn't find anything worthwhile.
Using BJTs could work, keeping things mostly discrete, and they do cause distortion, not sure how much in comparison with FETs, but I've been exploring this and it seems overly complex and not so easy to keep linear.
So I don't see any other viable solution besides trying the vactrols anyway...
Any ideas?
Based on my experience, VI limiters are a bad idea. The problem is that they are unstable, particularly under inductive loads. A VI limited amp connected to a transformer makes a relaxation oscillator; very bad for amp and speakers. The same thing can happen with some very reactive speaker loads.
Granted, new 2n3055's are a lot better than the original parts, but how old are your 2n3055s? It appears that on-semi has dropped the original 2n3055 and now sells what they used to call mj3055 as 2n3055. But why waste your time using these old parts when much better parts like 2SC5200 are available for reasonable prices?
Granted, new 2n3055's are a lot better than the original parts, but how old are your 2n3055s? It appears that on-semi has dropped the original 2n3055 and now sells what they used to call mj3055 as 2n3055. But why waste your time using these old parts when much better parts like 2SC5200 are available for reasonable prices?
Granted, those vi limiters do cause a lot of havoc when they act, and they're not even always 100% effective, which is the case on this amp project.
But what do you propose using instead? What would you use to keep the transistors within their SOA?
Not so many options available unfortunately, and crowbars or other nasty things like that are out of the question, actually even worse than the vi limiters.
The conditions of Vce and Ic on the outputs and drivers should be watched in any case, and if the action isn't applied between the vas and drivers as always in those vi limiters, then their action would need to be applied some other way.
I thought about using them to apply input compression on the amp, but how to be sure this action does keep within the SOA???
And if doing this, we're still looking at a proper input compression scheme, and there aren't that many that don't cause too much distortion.
What I've been exploring lately is BJT based, and it can work, and does cause more distortion when acting than the simple vactrols, although not sure about how they compare with JFETs... What I've tried lately can remain mostly transparent until action is required, then distortion creeps in as soon as they start acting. I suppose in a way we could live with that, because when they act, something much worse is going on, which would cause far more distortion and the limiting would fight against things getting worse.
With a proper time constant, a pumping effect could be avoided, but in any case, action would be audible, and obviously mixed up with the other cause of the problem (output short, overly hard load, overdriving, clipping...)
There are many of us in possession of 3055s and the likes. Old stock, new parts that were never used, or just people who are in areas on this planet where such parts are easy to find and cheap while the better newer and more expensive ones aren't that easily procured.
Plus the point of doing 3055 project is to push the limits of what can be done and reach respectable levels of performance out of those old trusty parts.
The bad thing is when buying such parts, there is a fair chance to get taken and end up with cheap faulty knock offs...
Even if it wasn't for this 3055 amp project, the search for a good input limiter is still worthwhile, and possible better alternatives to the vi limiters can also be explored.
But I'm not in favor of anything that shuts things down, like muting 100% in any way, when a non life threatening thing happens to the amp. Only the speakers' safety command such radical action. Like if DC is present at the amp output, there is no question that the speakers should be immediately disconnected, as fast as possible.
If it's a matter of DC at the amp output, there is a good chance the amp is what's wrong, perhaps blown or whatever, so first thing has to be disconnecting the speakers (SSR on the output), possibly muting the amp's input, just in case, and if some damage can be limited, using SSRs on the rails to shut the amp down completely.
If you have a viable alternative to the vi limiter, then why not share?
Interesting, I've been playing with VI limiter simulation today as it happens. I think I've a fairly workable solution to two issues with them, the tendency to oscillate, and the effect they can have on linearity some way below limiting conditions due to the gradual turn on of the transistor:
This is a dual slope version using a zener for the knee. The crucial components: Firstly R1, which means that Q1 can conduct upto a mA or two before the diode conducts, so that distortion is unaffected below this point.
Secondly C1 acts to damp the tendency to oscillation once D1 is conducting, but without any effect before limiting (assuming D1 is low capacitance).
This is a dual slope version using a zener for the knee. The crucial components: Firstly R1, which means that Q1 can conduct upto a mA or two before the diode conducts, so that distortion is unaffected below this point.
Secondly C1 acts to damp the tendency to oscillation once D1 is conducting, but without any effect before limiting (assuming D1 is low capacitance).
Last edited:
That tendency to oscillate is inherent to how the limiter functions, so although we can endeavor to minimize those, they can never be completely eliminated.
Having the time constant helps dampen those somewhat, and a compromise between the speedup cap placed on one of the sense resistors and the time constant one can be found that makes it kindof "acceptable", but it's always a compromise and there is no way around having all kinds of funny business going on when the limiter is acting.
They do start having some small effect long before they actually act. No matter how adamant M.K is about that fact, they are almost never totally transparent, as the transistors don't have a very sharp knee (vbe) and they do act a little way before the calculated values. If the knee was very sharp, that would cause issues as well anyway, although at the very last moment.
If the locus can be pushed far enough above the legit soa usage, then this small disturbance from premature action can be minimize, but that automatically means the output stage must be overbuilt to allow this, which pretty much defeats the need to use more than a single slope and if the output stage is really way over built, then a simple current limit could suffice.
But in any case, any foldback action to suppress the drive to the output stage will always cause serious disturbance, and the fact that the limiter is really fighting hard against the amp feedback makes it sometimes hard to act enough.
That's the main issue on our 3055 grounded bridge design, the vi limiter has the hardest time steeling enough drive and when it's fully activated, there is still enough drive to push the output stage over the soa.
This is why I'm also searching for a good combination that will help keep things under control, within soa, and hopefully remain as transparent as possible.
I've never tried this. Seems interesting. I've tried many things as well.
I've tried using schottky diodes, hoping the limiter could clamp a bit more drive away because of the lower forward drop, but that doesn't make a meaningful difference.
One thing that did help shunt a bunch more drive away was to use cfp instead of the singleton for the limiter. That way it is more capable and can sink more drive current from the vas, so less of it makes it to the output stage. But this still isn't quite sufficient to enforce the locus fully and keep everything within soa. It's much better, but I don't think it's enough.
Maybe it's the nature of the grounded bridge that makes it harder for the limiter to act effectively. At least compared to my earliest attempts, this is now working, although not quite enough.
The locus we calculated for recently, with the triple slope, on the 4 pairs output stage, has shown to be rather transparent all the way up to full power before clipping, until it acts, and all hell breaks loose then, but that's to be expected.
So it's calculated right, and works fairly well, but not enough for full soa compliance.
A combination with an input limiter will definitely help with this and make the whole thing work. But I'm seeking a good limiter with as little distortion as possible. I tried with bjts, but this gets a little complex with all discrete. And the bjts do cause some distortion.
It looks like no matter what, the vactrol solution remains the best one, and if those can't be obtained, then making them would be the last resort solution, but I think worth doing.
Having looked at many vactrol datasheets, I think the best suited would be the VTL5C9, which I think is pretty much gone from the market now.
I will be doing sims with vactrols now, and this is probably what will work best.
Such an input limiter wouldn't be used only when the vi limiter is acting, as it can also be made to alleviate the clipping effects, and in some other cases of overdrive or whatever that causes the amp's input stage to go out of balance.
I'm trying to figure out the input limiter using a vactrol and I don't have a good working model of the type I think would be suitable for this.
I have a model for the VTL5C2/2, which is incomplete, but this isn't that straightforward to alter to fit a different set of characteristics. That one has an overly high Ron with a high led current, and the Roff isn't very high compared to the VTL5C9, which I think is much better for this, because it gives it a much bigger dynamic range, while having a much lower Ron and at a lower led current.
Seems this type might have become almost impossible to obtain anyway, but the goal would be to make one with as similar characteristics as possible.
The 5C9 has an attack time of some 4ms and decay 50ms, which is pretty much in line with what we want for an input limiter.
This would prevent an overly complex discrete circuit.
Would anyone have a working model of the VTL5C9 vactrol?
And if anyone has been playing with making DIY vactrols, what combination of led/ldr would be best and come close enough to what the 5C9 does.
The dark Roff for the 5C9 is some 50Meg, so that's very nice, and the Ron can drop to about 100 with the highest led current.
I think a high intensity led could be a good way to push the ldr to a low enough Ron. The best led would be the one giving out the highest intensity at the ldr's peak sensitivity in the light bandwidth, if at all possible.
Since it would be DIY, then nothing prevents making a vactrol with 2 leds, to push the ldr as much as possible for the lowest Ron.
I have a model for the VTL5C2/2, which is incomplete, but this isn't that straightforward to alter to fit a different set of characteristics. That one has an overly high Ron with a high led current, and the Roff isn't very high compared to the VTL5C9, which I think is much better for this, because it gives it a much bigger dynamic range, while having a much lower Ron and at a lower led current.
Seems this type might have become almost impossible to obtain anyway, but the goal would be to make one with as similar characteristics as possible.
The 5C9 has an attack time of some 4ms and decay 50ms, which is pretty much in line with what we want for an input limiter.
This would prevent an overly complex discrete circuit.
Would anyone have a working model of the VTL5C9 vactrol?
And if anyone has been playing with making DIY vactrols, what combination of led/ldr would be best and come close enough to what the 5C9 does.
The dark Roff for the 5C9 is some 50Meg, so that's very nice, and the Ron can drop to about 100 with the highest led current.
I think a high intensity led could be a good way to push the ldr to a low enough Ron. The best led would be the one giving out the highest intensity at the ldr's peak sensitivity in the light bandwidth, if at all possible.
Since it would be DIY, then nothing prevents making a vactrol with 2 leds, to push the ldr as much as possible for the lowest Ron.
Years ago (~1984) I installed an LDR limiter on the monitor amp in a roller rink because the DJs were burning out the monitor speakers. The LDRs were something we had in stock and I filed the (red) LED domes flat (two channels). There were no high intensity LEDs back then. The circuit was very simple. A series input resistor about 10K shunted by the LDR. A bridge and resistors for the LED. I don't remember the exact values but I probably experimented to define them. The effect was very subtle. The volume simply stopped getting any louder beyond about 100dBA. The time constants of the LDR as a shunt are almost perfect for such a limiter, fast attach and a slow release.
We also built a limiter based on JFETs. It was so fast it activated on the first pulse above the threshold, distorting that single wave only, which is almost imperceptible to the listener. With JFETs as variable resistance, you have to add half the audio amplitude to the gate voltage making them ~symmetric. But even then there is a limit to the ~flat part of the curve so they do cause distortion at higher levels, but not a problem at microphone levels. In any case, as a protection device, there is ~no distortion until they are activated. My favorite configuration was a P-channel JFET driven from the collectors of two NPNs dumping the time constant cap. A friend of mine used CMOS 4000 logic chips.
My favorite solution for power amps is spend some money on silicon and use current limit only. That was the solution used by BGW. Today silicon is cheap. And if you don't want to spend money on a large heat sink, go class-D. Class-D protection can be a fuse. If you want to use up old 2n3055s, make some nice LITTLE amps.
We also built a limiter based on JFETs. It was so fast it activated on the first pulse above the threshold, distorting that single wave only, which is almost imperceptible to the listener. With JFETs as variable resistance, you have to add half the audio amplitude to the gate voltage making them ~symmetric. But even then there is a limit to the ~flat part of the curve so they do cause distortion at higher levels, but not a problem at microphone levels. In any case, as a protection device, there is ~no distortion until they are activated. My favorite configuration was a P-channel JFET driven from the collectors of two NPNs dumping the time constant cap. A friend of mine used CMOS 4000 logic chips.
My favorite solution for power amps is spend some money on silicon and use current limit only. That was the solution used by BGW. Today silicon is cheap. And if you don't want to spend money on a large heat sink, go class-D. Class-D protection can be a fuse. If you want to use up old 2n3055s, make some nice LITTLE amps.
Here is an interesting led that might be good when coupled with the right ldr:
https://eu.mouser.com/datasheet/2/239/Lite-On_LTL2P3TGZ2KS-032A-1175196.pdf
https://eu.mouser.com/datasheet/2/239/Lite-On_LTL2P3TGZ2KS-032A-1175196.pdf
But current limit only doesn't protect output devices efficiently.
Typical device(*) might need limiting to 4A for an output short where 50V is across the device, but only 1A should the opposite output device fail shorted (100V case - secondary breakdown is a real inconvenience here as without a very stringent limit the failure will simply cascade to other device and vaporize something). Same output device could happily power resistive 4 ohm load peaking at 12A without a limiter.
So just a current limiter means either using 3 output devices for 12A, and having no protection against output-short-to-supply-rail, or using 12 output devices for 12A which is crazy.
The secondary breakdown is a tough thing to guard against without good VI slope response, and in practice this means highly reactive loads will risk the SOA even at very modest currents.
(*) example is NJW1302 in fact, although there only sees to be a 1 second line, not a DC line.
There's a whole question of what consitutes dynamic SOA when handling audio signals in class B, as typical conditions are intermittent and peaky rather than continuous, so arguably a good limiter handles dyamically changing load as intelligently as it handles voltage and current. In fact this might be an area for microcontroller intervention ideally, which can thermally model the device including secondary breakdown to inform an input limiter, as well as spot gross faults as fast as possible.
Typical device(*) might need limiting to 4A for an output short where 50V is across the device, but only 1A should the opposite output device fail shorted (100V case - secondary breakdown is a real inconvenience here as without a very stringent limit the failure will simply cascade to other device and vaporize something). Same output device could happily power resistive 4 ohm load peaking at 12A without a limiter.
So just a current limiter means either using 3 output devices for 12A, and having no protection against output-short-to-supply-rail, or using 12 output devices for 12A which is crazy.
The secondary breakdown is a tough thing to guard against without good VI slope response, and in practice this means highly reactive loads will risk the SOA even at very modest currents.
(*) example is NJW1302 in fact, although there only sees to be a 1 second line, not a DC line.
There's a whole question of what consitutes dynamic SOA when handling audio signals in class B, as typical conditions are intermittent and peaky rather than continuous, so arguably a good limiter handles dyamically changing load as intelligently as it handles voltage and current. In fact this might be an area for microcontroller intervention ideally, which can thermally model the device including secondary breakdown to inform an input limiter, as well as spot gross faults as fast as possible.
So long as its inside the feedback loop, fuse distortion is significant.
Last edited:
I would definitely like to investigate the microcontroller avenue to command an input limiter. Perhaps not for this 3055 amp project, but for others.
Anything that would allow removing the vi limiter acting on the output stage drive signal, which is inside the feedback loop, would be quite welcome, and this would also do away with all that freaky funny business that the vi limiters cause when they're acting.
When they're acting, the vi limiters definitely cause a lot of distortion (and oscillations) and this can't be all good for the speakers in any case.
So if the remedy is acting by compressing the input drive signal, not only a lot of distortion would be avoided, which would be better for speakers, but it would also be easier on the ears as well.
This sounds like a very interesting track to follow up on. Maybe a good way to make use of arduinos, which can be procured and programmed by anyone anywhere on the planet, even in the poorest countries I'm sure an arduino isn't hard to get.
And about inserting fuses inside a feedback loop, this isn't at all a good idea, as when the fuse blows, the amp goes in open loop, and watch out for some freaky stuff to happen then!
Anyway, if we're designing a decent amp, I think it's important to do it right as much as humanly possible, and build in all the proper safeguards, so the amp doesn't go poof without warning. Especially a high power amp, I think should be built to be as bullet proof as possible, and be ready to face any abuse thrown at it.
And never forgetting this is DIY, so nothing too hard to make and parts should be easy to source anywhere, as much as possible. That's why I wouldn't consider anything SMT, and no exotic hard to find parts.
Anything that would allow removing the vi limiter acting on the output stage drive signal, which is inside the feedback loop, would be quite welcome, and this would also do away with all that freaky funny business that the vi limiters cause when they're acting.
When they're acting, the vi limiters definitely cause a lot of distortion (and oscillations) and this can't be all good for the speakers in any case.
So if the remedy is acting by compressing the input drive signal, not only a lot of distortion would be avoided, which would be better for speakers, but it would also be easier on the ears as well.
This sounds like a very interesting track to follow up on. Maybe a good way to make use of arduinos, which can be procured and programmed by anyone anywhere on the planet, even in the poorest countries I'm sure an arduino isn't hard to get.
And about inserting fuses inside a feedback loop, this isn't at all a good idea, as when the fuse blows, the amp goes in open loop, and watch out for some freaky stuff to happen then!
Anyway, if we're designing a decent amp, I think it's important to do it right as much as humanly possible, and build in all the proper safeguards, so the amp doesn't go poof without warning. Especially a high power amp, I think should be built to be as bullet proof as possible, and be ready to face any abuse thrown at it.
And never forgetting this is DIY, so nothing too hard to make and parts should be easy to source anywhere, as much as possible. That's why I wouldn't consider anything SMT, and no exotic hard to find parts.
Perhaps a hybrid approach where a limiter circuit just activates an opto coupler, and another opto coupler removes drive from the drivers under remote command. However there is the timeliness issue - on a dead short the tranditional limiter transistor will switch on in a fraction of a microsecond, limiting the build up of current that might otherwise only be limited by the large filter caps and wiring.
Output device with built-in SOA monitoring would be the best approach, after all linear regulators already have this in their over-current and over-temperature cutout circuitry. A bomb-proof output transistor would be a popular replacement device for sure.
Output device with built-in SOA monitoring would be the best approach, after all linear regulators already have this in their over-current and over-temperature cutout circuitry. A bomb-proof output transistor would be a popular replacement device for sure.
I have used a two-pronged approach in one of my designs. A high current limiter prevents instantaneous destruction while an IV limiter opens a loudspeaker protection relay/overload cut out. The "instantaneous" limiter needs to protect the transistors for approx, the 1ms relay response time. Also use a fast response relay. Relay then has to be reset to operate once again, supposedly after checking and fixing a short.
Main reason is that the IV limiter is sort of passive, does not increase distortion (not as much as if the transistors operate internally to the amp. anyway).
Main reason is that the IV limiter is sort of passive, does not increase distortion (not as much as if the transistors operate internally to the amp. anyway).
That solution works fine for class AB and H. Adding silicon to boost SOA on class D is anything but straightforward. You end up needing to make something electrically large in terms of fractions of a wavelength (and heat sink it). They just don’t make 200 amp transistors with the parasitics of a surface mount 2N3904, in a package you can get the heat out of. Paralleling up devices isn’t so easy when you need to hit them all with a coherent wavefront - charging and discharging ever growing capacitances. These amps aren’t going to complain much about adding more 3055’s if you need them. I’ve seen a dozen or more 3773’s in parallel used on old Crowns and they work just fine.
BTW, 200 watts per channel is considered “a little amp”.
I have never had any serious problems with dual slope VI limiters with a single time constant to keep oscillation in check. It slows it down, but you really DO have a nearly square SOA at 100 microseconds. For that matter, I usually use the 100 millisecond SOA as my limiting threshold, but keep the zero crossing (short circuit ) current inside DC SOA. Because shorted speaker wires can linger until somebody notices. I typically limit my peak VAS currents to around 30mA and a single high gain (hFE 200) transistor has no problem shutting it down. If you need more than 20 mA out of your VAS you need higher gain drivers/predrivers.
The real problem with using these limiters on a grounded bridge isn’t inherent to the limiters. They would work fine on a single non-bridge, or both halves of an independently driven bridge. Shutting down both hslves of the amp simultaneously is where the problem lies.
Last edited:
I have used PIC micro's extensively for speaker protection.
I have a simple DC detect with a relay and another version with a solid state relay.
I also designed a DC protect, power up thump delay, mains fail anti thump and over current protect speaker protect circuit.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
One thing would be good to do, which would be to determine when there is a short on the amp output, as opposed to other things less alarming like clipping or some other overloads due to complex load or overly low impedance.
In the case of anything other than a dead short, we don't want to shut the amp down, just limit things to keep it soa safe.
But when there is an actual short on the output, the amp could be shut down, decoupled from the speakers (SSR relay) while the vi limiter just tries to keep things from going too far.
It's hard to electronically distinguish between a dead short and anything else that requires limiting action. The actual action required isn't the same if it's a real short.
I think a bit of microcontroller housekeeping might be good to add to the pure analog stuff to make it more of a complete solution.
I've been trying to find a good input muting solution that could be used as an input limiter, but there aren't that many good ways to do this, and so far, the best way that works great with the lowest distortion introduced has been the vactrol, but now it's getting so difficult to get those things, they're becoming unobtainium material.
I was thinking the best vactrol type to suit the bill was the VTL5C9, for its inherent fairly fast attack and natural slow release, which makes for simpler command circuits, needing hardly any time constants, and that vactrol's slow release time is pretty well suited for audio to prevent any pumping effect. And it has the highest (that I know of) dynamic range, with pretty high Roff and fairly low Ron.
However, as a full muting circuit, despite its high dynamic range, it may not be "full enough muting".
Still, as a simple and effective compressor, it really would be nice to use, if we could still obtain them.
So I've been digging into other solutions, including the BJT approach, but that really opens a dang can of worms and it's not as good.
Now the JFET solutions has been used by many out there, and may just be the last doable solution that may not work quite as well as a vactrol, but better than BJTs and whatever else.
The thing is, I'm not sure how this can work well enough for a balanced input, keeping things balanced, not reverting back and forth to unbalanced.
I've done some simulations, but haven't yet found anything that really is satisfying enough.
In any case, I'm pretty sure the combined use of a proper input limiter along with the vi limiter, should provide adequate protection, hopefully to keep the outputs SOA safe, even in this grounded bridge topology which seems to be more difficult to control by a vi limiter.
When I first tried adding a limiter only to the high side, thinking the low side just "follows" the high side kind of like a slave, I quickly found out that would never work. This grounded bridge is really symmetric and high side and low side both work the same, without one following the other, they just do the exact same thing in opposition direction.
So, when I realized that and tried a proper limiter on both sides, things looked a lot better and seemed to be feasible. And it's getting pretty close, however I find it insufficient, as it's not effective enough to "keep off the grass" (outside the SOA).
Adding CFP to the limiter helped a lot in making it more forceful to get more of the activity outside the SOA brought back in it. But there is still too much going on outside of the SOA to be really safe, especially on a dead short. With an input limiter acting fairly fast, whatever goes beyond the SOA would remain in short duration, when the input compression comes into play and further decreases the drive signal, then I think this might just do the job, possibly even on a dead short. And if there is a way to make the difference between a short and anything else, then the speakers can be disconnected in addition to everything else.
Obviously there would be no such additional protection if whatever shorted the output happens to remain once the speakers are disconnected. Like for example if the short isn't happening beyond the output speaker plug.
I wonder how a PIC microcontroller would be doable by some people out there.
I think the arduino might be better for this, as there is a free software download, that works on any platform, with a usb connection to any computer, allowing anyone anywhere to quickly program for example the arduino UNO, then take the chip off the UNO to put it in a socket on a pcb for the amp housekeeping system.
I've tried collecting a few odds and ends, microcontroller wise, to be able to program those things, from a mac, and I haven't yet done anything with PICs. But I've played a bit with arduinos, and that works nicely on mac.
If we design something that any DIYer out there, anywhere on this planet, can make, it should really be doable by anyone. So arduino may be a better choice for this. And it's simple to do. There isn't even any need for a DIYer to know much at all about programing to do this, only download the software, use the provided source, and with the arduino usb connected, just upload the code to it, and there it is. Prove me wrong please, about the PIC. I have a few tiny PICs, and a couple of odds and ends, but nothing yet that worked as well on a mac as an arduino.
In the case of anything other than a dead short, we don't want to shut the amp down, just limit things to keep it soa safe.
But when there is an actual short on the output, the amp could be shut down, decoupled from the speakers (SSR relay) while the vi limiter just tries to keep things from going too far.
It's hard to electronically distinguish between a dead short and anything else that requires limiting action. The actual action required isn't the same if it's a real short.
I think a bit of microcontroller housekeeping might be good to add to the pure analog stuff to make it more of a complete solution.
I've been trying to find a good input muting solution that could be used as an input limiter, but there aren't that many good ways to do this, and so far, the best way that works great with the lowest distortion introduced has been the vactrol, but now it's getting so difficult to get those things, they're becoming unobtainium material.
I was thinking the best vactrol type to suit the bill was the VTL5C9, for its inherent fairly fast attack and natural slow release, which makes for simpler command circuits, needing hardly any time constants, and that vactrol's slow release time is pretty well suited for audio to prevent any pumping effect. And it has the highest (that I know of) dynamic range, with pretty high Roff and fairly low Ron.
However, as a full muting circuit, despite its high dynamic range, it may not be "full enough muting".
Still, as a simple and effective compressor, it really would be nice to use, if we could still obtain them.
So I've been digging into other solutions, including the BJT approach, but that really opens a dang can of worms and it's not as good.
Now the JFET solutions has been used by many out there, and may just be the last doable solution that may not work quite as well as a vactrol, but better than BJTs and whatever else.
The thing is, I'm not sure how this can work well enough for a balanced input, keeping things balanced, not reverting back and forth to unbalanced.
I've done some simulations, but haven't yet found anything that really is satisfying enough.
In any case, I'm pretty sure the combined use of a proper input limiter along with the vi limiter, should provide adequate protection, hopefully to keep the outputs SOA safe, even in this grounded bridge topology which seems to be more difficult to control by a vi limiter.
When I first tried adding a limiter only to the high side, thinking the low side just "follows" the high side kind of like a slave, I quickly found out that would never work. This grounded bridge is really symmetric and high side and low side both work the same, without one following the other, they just do the exact same thing in opposition direction.
So, when I realized that and tried a proper limiter on both sides, things looked a lot better and seemed to be feasible. And it's getting pretty close, however I find it insufficient, as it's not effective enough to "keep off the grass" (outside the SOA).
Adding CFP to the limiter helped a lot in making it more forceful to get more of the activity outside the SOA brought back in it. But there is still too much going on outside of the SOA to be really safe, especially on a dead short. With an input limiter acting fairly fast, whatever goes beyond the SOA would remain in short duration, when the input compression comes into play and further decreases the drive signal, then I think this might just do the job, possibly even on a dead short. And if there is a way to make the difference between a short and anything else, then the speakers can be disconnected in addition to everything else.
Obviously there would be no such additional protection if whatever shorted the output happens to remain once the speakers are disconnected. Like for example if the short isn't happening beyond the output speaker plug.
I wonder how a PIC microcontroller would be doable by some people out there.
I think the arduino might be better for this, as there is a free software download, that works on any platform, with a usb connection to any computer, allowing anyone anywhere to quickly program for example the arduino UNO, then take the chip off the UNO to put it in a socket on a pcb for the amp housekeeping system.
I've tried collecting a few odds and ends, microcontroller wise, to be able to program those things, from a mac, and I haven't yet done anything with PICs. But I've played a bit with arduinos, and that works nicely on mac.
If we design something that any DIYer out there, anywhere on this planet, can make, it should really be doable by anyone. So arduino may be a better choice for this. And it's simple to do. There isn't even any need for a DIYer to know much at all about programing to do this, only download the software, use the provided source, and with the arduino usb connected, just upload the code to it, and there it is. Prove me wrong please, about the PIC. I have a few tiny PICs, and a couple of odds and ends, but nothing yet that worked as well on a mac as an arduino.
There is one more thing I'd like to know about that needs to be done on this grounded bridge amp design:
How to properly use a Tian probe to check for margins?
I've tried but didn't get conclusive results, and this I think is just due to the fact that there isn't just a single feedback loop. Being a grounded bridge, entirely symmetrical, with each side having its own feedback loop and compensation, I don't see how a Tian probe can be made to work.
Someone could explain this please?
How to properly use a Tian probe to check for margins?
I've tried but didn't get conclusive results, and this I think is just due to the fact that there isn't just a single feedback loop. Being a grounded bridge, entirely symmetrical, with each side having its own feedback loop and compensation, I don't see how a Tian probe can be made to work.
Someone could explain this please?
Browsing through the OPA1632 datasheet, I just realized something I didn't pay attention to before. That thing has a built-in shutdown input, and that could be used for a full shutdown/muting in extreme cases, such as a short. Shutting down both the amp's input drive along with the output to the speaker with a SSR would be a very good thing to do.