Back to DC protection: Detection Circuit details...
Ok...so a large DC offset is detected....and the MOSFET switches, which were previously turned on by default at power up, are promptly turned off...
The supply rails collapse to zero volts, (assuming these are returned to ground by large-value resistors), and the offset detection circuit resets.....
Consequently, MOSFETS are switched back on, and the cycle resumes....
Rapid...off...on...off....on...untill the speakers succumb, or the amp. is manually turned off....
I suggest the detection circuit has to be latching.....viz: ounce triggered, it must keep the MOSFETs off untill the amp. is manually turned off...and on again...
Ideas and analysis welcome...
Ok...so a large DC offset is detected....and the MOSFET switches, which were previously turned on by default at power up, are promptly turned off...
The supply rails collapse to zero volts, (assuming these are returned to ground by large-value resistors), and the offset detection circuit resets.....
Consequently, MOSFETS are switched back on, and the cycle resumes....
Rapid...off...on...off....on...untill the speakers succumb, or the amp. is manually turned off....
I suggest the detection circuit has to be latching.....viz: ounce triggered, it must keep the MOSFETs off untill the amp. is manually turned off...and on again...
Ideas and analysis welcome...
Better than a latch
It is probably a better idea a long timeout before automatic retry much as for example regular switching power supplies do.
This way there is no need for manual resetting. The sense circuit must be fast enough to operate in the sub-millisecond range to avoid speaker damage.
The same power cutoff circuitry can then be applied for transient overload protection (load shorts / misbehaving at certain frequencies).
Rodolfo
It is probably a better idea a long timeout before automatic retry much as for example regular switching power supplies do.
This way there is no need for manual resetting. The sense circuit must be fast enough to operate in the sub-millisecond range to avoid speaker damage.
The same power cutoff circuitry can then be applied for transient overload protection (load shorts / misbehaving at certain frequencies).
Rodolfo
anatech said:
AHHH yes post #11...missed that...he's absolutely right...Cheers..:
http://www.diyaudio.com/forums/showthread.php?postid=580607#post580607
Ok, some restatement of objectives may be in order.
The motivation for the shut-off-the-power-rails idea is that there are faults in output stages that cannot be stopped or even ameliorated by messing with the output devices. Most notably there are occasions where the output devices themselves have died, and are no longer capable of comprehending orders from driver transistors or protection circuits. When this scenario happens, the full power supply is dumped into the speakers, and the speakers are magically transformed into smoke generators, at some considerable loss in their audio capabilities.
We detect this situation by looking for significant DC levels being delivered to the speaker outputs, with all due concern paid to not tripping on minor DC offsets or low frequency program content.
With such a detection in place, we have to decide what to do with the knowledge.
There are really only two things we can do. Well, OK, three if you count doing nothing and letting the speaker fry. But we're not going to do that. So:
1) we can open the path from the power supply to the speaker by inserting some large impedance
2) we can shunt the speaker with a low impedance to eat the DC power and keep the majority of it out of the speakers.
(2) is really only a short term solution, given the really large amounts of power available from the AC mains powered DC power supply. All the (2) ways to protect speakers I can think of eventually resolve into a (1). Triacs and anti-parallel SCRs work for as long as it takes to open either a DC rail fuse, an AC mains fuse, or a wall-power breaker. I don't think I could come up with a shunt scheme that would dissipate enough power to just keep shunting the DC supply in a non-damaging manner.
(1) schemes take not of the idea that if you have to open something anyway, why not do that first. Hence the relay solution - insert a single mechanical contact in series with the speaker, and you're done. It works fine in many cases, and has for decades. However, relays are electromechanical, and prone to all that their ancestry implies. MOSFETs are just another way to do the same, but at a point that's outside the delicate normal operation of the audio path. Same motivation, different placement.
I've tried a lot, and I can't really come up with a scenario where the amplifier, once it's generated a DC fault condition I'd like to protect which would self-heal enough for me to trust it until later, and as a result, I can't think of a good reason not to have the detection circuit latch. Some possibilities I came up with were thermal overloading, which went away when the amp cooled and transient overloads of the front end or drivers.
I'd handle the thermal situations more directly, perhaps with thermal sensors (something like Vbe multipliers or thermistors) watching the output devices, drivers, heatsinks and general air temperature and linearized/sensed by a $2.00 PIC which decides that there's a problem.
I can't come up with any scenario where the amp self-heals that doesn't also involve possible death if I turn the amp back on. So, in my mind at least, of course the fault detect must be latching. As I said at least once before, the design of the fault detector is challenging to get it right. I'd put my efforts into getting a detector to really detect failure conditions, but if there is any question about whether an output is shorted or not, I want the speakers isolated from the DC power until I can sort things out.
Wait-a-while kinds of protection circuits are OK, but I think that requiring a human to cycle the AC is a reasonable requirement for possible speaker-frying incidents.
The motivation for the shut-off-the-power-rails idea is that there are faults in output stages that cannot be stopped or even ameliorated by messing with the output devices. Most notably there are occasions where the output devices themselves have died, and are no longer capable of comprehending orders from driver transistors or protection circuits. When this scenario happens, the full power supply is dumped into the speakers, and the speakers are magically transformed into smoke generators, at some considerable loss in their audio capabilities.
We detect this situation by looking for significant DC levels being delivered to the speaker outputs, with all due concern paid to not tripping on minor DC offsets or low frequency program content.
With such a detection in place, we have to decide what to do with the knowledge.
There are really only two things we can do. Well, OK, three if you count doing nothing and letting the speaker fry. But we're not going to do that. So:
1) we can open the path from the power supply to the speaker by inserting some large impedance
2) we can shunt the speaker with a low impedance to eat the DC power and keep the majority of it out of the speakers.
(2) is really only a short term solution, given the really large amounts of power available from the AC mains powered DC power supply. All the (2) ways to protect speakers I can think of eventually resolve into a (1). Triacs and anti-parallel SCRs work for as long as it takes to open either a DC rail fuse, an AC mains fuse, or a wall-power breaker. I don't think I could come up with a shunt scheme that would dissipate enough power to just keep shunting the DC supply in a non-damaging manner.
(1) schemes take not of the idea that if you have to open something anyway, why not do that first. Hence the relay solution - insert a single mechanical contact in series with the speaker, and you're done. It works fine in many cases, and has for decades. However, relays are electromechanical, and prone to all that their ancestry implies. MOSFETs are just another way to do the same, but at a point that's outside the delicate normal operation of the audio path. Same motivation, different placement.
I've tried a lot, and I can't really come up with a scenario where the amplifier, once it's generated a DC fault condition I'd like to protect which would self-heal enough for me to trust it until later, and as a result, I can't think of a good reason not to have the detection circuit latch. Some possibilities I came up with were thermal overloading, which went away when the amp cooled and transient overloads of the front end or drivers.
I'd handle the thermal situations more directly, perhaps with thermal sensors (something like Vbe multipliers or thermistors) watching the output devices, drivers, heatsinks and general air temperature and linearized/sensed by a $2.00 PIC which decides that there's a problem.
I can't come up with any scenario where the amp self-heals that doesn't also involve possible death if I turn the amp back on. So, in my mind at least, of course the fault detect must be latching. As I said at least once before, the design of the fault detector is challenging to get it right. I'd put my efforts into getting a detector to really detect failure conditions, but if there is any question about whether an output is shorted or not, I want the speakers isolated from the DC power until I can sort things out.
Wait-a-while kinds of protection circuits are OK, but I think that requiring a human to cycle the AC is a reasonable requirement for possible speaker-frying incidents.
Hi R.G.,
Thermal switches are tried and true. Bolt one to each heatsink.
Power cycling will teach some not to play with speaker wires while the amp is running, or load it down with as many speakers you can find. I like to inconvenience abusers slightly. Makes them think sometimes. Do the same for thermal trip (for the same reasons).
-Chris
Thermal switches are tried and true. Bolt one to each heatsink.
Power cycling will teach some not to play with speaker wires while the amp is running, or load it down with as many speakers you can find. I like to inconvenience abusers slightly. Makes them think sometimes. Do the same for thermal trip (for the same reasons).
-Chris
Yeah, they are. I've used them, and they're great at a simple level.
On the other hand, any cheap semiconductor sensor when coupled with some programmable logic lets you have lattitude to do... anything you like... with the info that temperature is rising.
F'rinstance: Use thermistors and a $1.58 (unit price, Digikey catalog) 8 pin 12C508 microcontroller and not only sense when to shut power down, but also (a) slowly ramp up a fan (b) turn on an overtemp light (c) finally put the amp into some kind of standby mode until things cool off, then bring it back up. Program it in Basic with a free compiler and an almost-zero-cost programmer.
Yeah, that's always fun. Use the PIC to set a special on-off-on-off rhythm so that only people who know the "code" can get the amp to come back up....
Hello,
I feel confused by the logic in here.
1) what's the point of having "sub-millisecond" switching in DC protection? DC sensing latency will take about 0.1 to 0.5s. Else, it's not "DC" but audio-range.
2) If one inserts a capacitor to protect a tweeter from DC in active x-over, then what's the point of having a DC output protection for it?
Conversely, one advantage of active x-over is to drive directly speakers. Isn't that advantage lost by inserting this capacitor (which will need to be a good/expensive one)?
If capacitor insertion is "mandatory to protect tweeter", why not use the same solution for all speakers
3) For latching issue, what about a fuse on each supply rail?
4) What is the targeted cost? Ok, output protection is important. But I tend to listen to music more often than test DC protection
BTW, I'm not sure that people designing modern relays will be happy reading that they do "ancestry things".
I feel confused by the logic in here.
1) what's the point of having "sub-millisecond" switching in DC protection? DC sensing latency will take about 0.1 to 0.5s. Else, it's not "DC" but audio-range.
2) If one inserts a capacitor to protect a tweeter from DC in active x-over, then what's the point of having a DC output protection for it?
Conversely, one advantage of active x-over is to drive directly speakers. Isn't that advantage lost by inserting this capacitor (which will need to be a good/expensive one)?
If capacitor insertion is "mandatory to protect tweeter", why not use the same solution for all speakers
3) For latching issue, what about a fuse on each supply rail?
4) What is the targeted cost? Ok, output protection is important. But I tend to listen to music more often than test DC protection
BTW, I'm not sure that people designing modern relays will be happy reading that they do "ancestry things".
Me, too, sometimes. It's OK, just hang in there and it'll pass.
Actually, the slowness of the DC sensing makes fast switching mandatory.
If you accept that some method is needed to protect speakers from an amplifier fault that will burn them up, and that an electronic means of turning the power off is necessary at all - and there are some people who do not, preferring music somehow unpolluted by components in the box not involved in audio - then you are immediately in the soup for how to do this.
There is no practical means other than doing some kind of DC sense. You have to distinguish DC faults from low frequency program material, and that necessarily means slow sensing. However, for the time that the sensing is going on, the speakers are heating. There is a finite time that a speaker will sit there being fed a solid voltage level before it burns out. It depends on the voltage being fed to it, of course, but there is a time limit where if you don't get the DC turned off, the speaker is damaged.
The more of that time that is eaten up by your DC sensing ( and the longer the sense time, the better job it does at real detection) then the shorter time left before the speaker dies. Necessarily then, the shorter the disconnect time, the better job you can do at detection. That's not to say that relays can't do it. The MOSFET trick is just one other way. The're faster than relays. That could potentially give you more time to do a better (i.e. slower) detection job.
Note that the problem is especially acute with multiamped tweeters, which don't have as large a thermal time constant as big, beefy woofers. I think that as a generality, they die quicker from DC.
There isn't any. This idea is not intended to save capacitor isolated speakers. If you have a good reliable cap, you can use it for DC protection.
Yes. So if you multi-amp and do not use a capacitor isolator, then you are either sensing and protecting, or ingenuously assuming that a failure will never happen. Sometimes you win at that. Sometimes you don't.
Good idea - except many people hate the sound of capacitor isolated speakers, particularly bass speakers. The capacitor cost and quality gets out of hand quickly, and some people can hear capacitor pollution in any sound stream, yes?
Good idea. But you gotta blow them both at the same time. If you don't blow them both, that guarantees that the unblown side will be dumping its DC into the speakers. Besides, I hate buying fuses. I like to have things quietly turn themselves off and be resettable. But that's just me.
What targetted cost? As several plies of this discussion have pointed out, it's not really all that expensive. Yes, if you were building a zillion of them, it would be smart to shave pennies, but the cost of some MOSFETs and some small signal detection and logic seems to be on the same order as the series relay that they were meant to replace.
I think a more pertinent question is - what's the targetted cost of replacing speakers versus the amount you're willing to spend for insurance? Most of us buy insurance against various disasters to our health and property when we can. Most of us never need the insurance. But still, the idea that we'd have to rebuild our houses or cars out of our pocket or get substandard medical help because we can't pay for good medical help is abhorrent enough to make us dig into our pockets for a little contribution every day. there's nothing that says anyone has to put protection into an amp against any fault. Some people take perfectly good SOA protection OUT of working amps on the theory that they'll never need it and that it subtly pollutes the sound. That's OK, too.
Me, I'm cheap enough that I'd rather spend some money once and keep my speakers.
I don't think it will bother them. Most people who design relays are aware that relays have a history that goes back to the days when electrical experimenters were arguing about the exponent to be attached to the current in the expression V = f(I)*R. Ohm's contribution to that debate, which cause a lot of stir, was that the exponent was unity.
Relays have a LONG history, and while modern relay design is not irrelevant, it is right up against the laws of physics, and I don't think any relay designer expects to get another doubling of performance. They're fighting for percentage points in performance, and I think they realize it. The spotlight of technology is not pointing at relays these days. People who pride themselves on being cutting edge don't exactly aim at relay design to change the world.
R.G. said:
Sorry, there are untold zillions of PC's, CRT's and TV's out there routinely pumping anywhere from 150 to over 300W - real nasty power -, out of which surely millions have failed catastrophicaly beyond the PSU level. Yet - excluding the blown fuse case which is not necessarily the single consequence - only seldom start a fire.
We know it is not easy neither interesting (at least for me) to delve into smart and user friendly protection schemes, I only want to point is it is both possible and cheap if done right.
And of course the PIC way is neat, but actual developement cannot be recommended uness considered by itself an educational experience.
Rodolfo
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.