Generally speaking, I fully agree with that. If I try to be consciencious, my next question is:
what's the time window for DC detection ? (sensing+acting)
Genrally, DC is extracted by an RC net. In order to avoid sensing error, corner frequency is chosen very low. Sensing error mainly come from audio residue not filtered enough. I think a 2nd order filter (RCRC is surely enough) is a simple but great improvement. It can give better attenuation (less sensing errors), and can have a higher corner frequency, and consequently a shorter latency for DC sensing.
Tweeter coils are more fragile for sure.
Sorry for the simple/silly questions about output capacitor.
I knew the answers, but I wanted to point out that backup solution, and have your opinion on it.
Which leads to:
cost of active output protection should not be higher than the price of a very good coupling capacitor providing the same protection
And of course, as pointed out elsewhere, output protection (active circuitry or coupling cap) should cost less than the speakers it protects.
This DC is the one that has been detected, so speaker is expected to be safe due to DC protection relay (elctromechanical or solid state, whatever). And until problem is solved, DC protection will protect speaker unless the other fuse blows off. Ok, this is not ideal, but an idea.
Another idea is to use supply monitoring. Since the main causes of DC fault discussed here do reflect on supply rails, then checking that rails are healthy can be all the more efficient.
It reminds me that some people tend to put a bigger fuse when previous one has burnt... Hum...
And a piece of PCB, some mounting screws, some connectors, maybe a separate supply (depending on latch circuitry), some inches of cable...
For example, in an amplifier, the output stage is rarely the most expensive stage.
Don't take me wrong. I'm not talking about relay or FET price, but keeping track of overall output protection cost. Illimited budget tend to result in "smart but out of the point" solutions.
Uh? These people are rich, or have found a cure for -component- aging 😀
Maybe no, maybe yes: HF and micro-mechanics are their new playfield... But that's another story.
Why not just detect whether the output resembles the input? Trip at some level of output does not equal input (scaled of course). This should be really fast. It does not matter whether a relay disconnects or Mosfets turn off (or short 😀 )
-Chris
-Chris
anatech said:
It is a good idea but does not cover the overload condition (ouput short that is not really 0 ohms), that in turn could lead to output device failure.
Requires separate protection, if it is allready in place, then fine.
Rodolfo
Hi Rodolfo,
I was considering the case of DC offset or severe clipping to shut things down. Over current is time delayed and detected elsewhere.
-Chris
I was considering the case of DC offset or severe clipping to shut things down. Over current is time delayed and detected elsewhere.
-Chris
Hi,
I'm jumping in here & correct me if I'm wrong.
Input to output difference checking will surely detect a very low output impedance. Reasoning:- The amp will start distorting and/or the gain will fall. Both of these can be detected and trigger the protection.
Similarly a DC just starting to develop on the output can be detected before it has even reached the rail voltage. The speed of activation and cut off then determine how long the error condition lasts.
A very fast protection system should be able to protect both the speaker (all varieties) and the remaining undamaged parts of the amp.
BTW I second the latching, I have promoted it on 4 recent threads and only 2 contributors have supported the idea. It seems that most users only have their music systems on while they listen with a finger over the off switch (manual latch) in case of internal mishaps.
I'm jumping in here & correct me if I'm wrong.
Input to output difference checking will surely detect a very low output impedance. Reasoning:- The amp will start distorting and/or the gain will fall. Both of these can be detected and trigger the protection.
Similarly a DC just starting to develop on the output can be detected before it has even reached the rail voltage. The speed of activation and cut off then determine how long the error condition lasts.
A very fast protection system should be able to protect both the speaker (all varieties) and the remaining undamaged parts of the amp.
BTW I second the latching, I have promoted it on 4 recent threads and only 2 contributors have supported the idea. It seems that most users only have their music systems on while they listen with a finger over the off switch (manual latch) in case of internal mishaps.
Hi Andrew,
I think the idea is that because of a few poorly designed protection circuits, they are all bad. Then the fanatical side takes over and many jump on the "pure circuit" bandwagon. No protection, no disconnect no hope.
I still love the answer an un-named fella at Bryston gave me a few years back ... "our amps never fail". He really said that.
-Chris
I think the idea is that because of a few poorly designed protection circuits, they are all bad. Then the fanatical side takes over and many jump on the "pure circuit" bandwagon. No protection, no disconnect no hope.
I still love the answer an un-named fella at Bryston gave me a few years back ... "our amps never fail". He really said that.
-Chris
General protection mumblings
The thread prompted me to delve a little more in ideas about something we usually treat as more of an afterthought, the issue of speaker protection, and catastrophic output stage damage prevention-securing.
Probably a good approach is first to define the scenario as to what we want to protect and of which electrical hazzard, which are the probable failure modes leading to it, and how to best sense them both reliably and economically.
Then to define a protection strategy - once the trigger signal is aquired - taking into account again reliability, economy and user friendliness.
Not pretending to give an authoritative answer, I was thinking on the following lines:
What to protect:
- Speakers, in the form of excessive instantaneous voice coil current, and excessive steady state current (AC and DC).
- Power output devices, in the form of excessive current and dissipation.
How to sense:
A cheap and handy place to sense is with a small value resistor (0.1 ohms or less) inserted between the power supply ground and the amplifier/speaker return ground.
An inexpensive OpAmp amplifies and rectifies full wave this signal, which is fed to one half of an equally inexpensive dual comparator. This one takes care of sending a digital signal whenever the instantaneous current - whichever the polarity - exceeds a defined threshold.
The second half is fed an integrated (RC filtered) copy of the amplified signal, providing again a digital fail signal when the average current (AC or DC) exceeds a second threshold. Usually outputs can be wire-ored.
Thermal excursions can equally be sensed with thermistors or diodes, detected and ored.
What to do:
MOSFET (or bipolars for that matter) in the supply rails, fed by independent floating supplies are a good idea as posted earlier. Other schemes like supply shutoff in the case of SMPS or regulated supplies are possible. Crowbar SCR's and fuses / breakers are a further option, but manual reset is required.
Which brings the issue of user friendliness. It is desirable and possible to latch the failure signal, and to retry with periods no smaller than a couple of seconds.
Accidental failures like shorts, or gross spikes generated by loose connectors or plugging - unplugging are taken care off by the fast acting protection, and no permanent high duty cycle damaging power is allowed to be fed to the protected devices.
Just my oppinion.
Rodolfo
The thread prompted me to delve a little more in ideas about something we usually treat as more of an afterthought, the issue of speaker protection, and catastrophic output stage damage prevention-securing.
Probably a good approach is first to define the scenario as to what we want to protect and of which electrical hazzard, which are the probable failure modes leading to it, and how to best sense them both reliably and economically.
Then to define a protection strategy - once the trigger signal is aquired - taking into account again reliability, economy and user friendliness.
Not pretending to give an authoritative answer, I was thinking on the following lines:
What to protect:
- Speakers, in the form of excessive instantaneous voice coil current, and excessive steady state current (AC and DC).
- Power output devices, in the form of excessive current and dissipation.
How to sense:
A cheap and handy place to sense is with a small value resistor (0.1 ohms or less) inserted between the power supply ground and the amplifier/speaker return ground.
An inexpensive OpAmp amplifies and rectifies full wave this signal, which is fed to one half of an equally inexpensive dual comparator. This one takes care of sending a digital signal whenever the instantaneous current - whichever the polarity - exceeds a defined threshold.
The second half is fed an integrated (RC filtered) copy of the amplified signal, providing again a digital fail signal when the average current (AC or DC) exceeds a second threshold. Usually outputs can be wire-ored.
Thermal excursions can equally be sensed with thermistors or diodes, detected and ored.
What to do:
MOSFET (or bipolars for that matter) in the supply rails, fed by independent floating supplies are a good idea as posted earlier. Other schemes like supply shutoff in the case of SMPS or regulated supplies are possible. Crowbar SCR's and fuses / breakers are a further option, but manual reset is required.
Which brings the issue of user friendliness. It is desirable and possible to latch the failure signal, and to retry with periods no smaller than a couple of seconds.
Accidental failures like shorts, or gross spikes generated by loose connectors or plugging - unplugging are taken care off by the fast acting protection, and no permanent high duty cycle damaging power is allowed to be fed to the protected devices.
Just my oppinion.
Rodolfo
Hi Rodolfo,
The series element will lower the damping factor. A current transformer may work. Just sense across the existing emitter / source resistors. Sense the deviation from the original signal however you want and act on it. Trip on thermal also.
Disconnect the supply from the load (switch the primary of the power transformer, short the output or open the connection to the speaker). My personal belief is the supply rails should be left alone.
Latch this condition until the power is cycled. This prevents the cycling of fault current. If the cause is user mis-intervention, then make the user mildly annoyed to correct that behaviour. Hopefully they didn't read the user manual and will pay an estimate charge.
Keep these circuits simple. Complicated circuits break and are tough to troubleshoot. I have repaired many protection circuits, many overly complicated. The amps were fine (of course).
-Chris
The series element will lower the damping factor. A current transformer may work. Just sense across the existing emitter / source resistors. Sense the deviation from the original signal however you want and act on it. Trip on thermal also.
Disconnect the supply from the load (switch the primary of the power transformer, short the output or open the connection to the speaker). My personal belief is the supply rails should be left alone.
Latch this condition until the power is cycled. This prevents the cycling of fault current. If the cause is user mis-intervention, then make the user mildly annoyed to correct that behaviour. Hopefully they didn't read the user manual and will pay an estimate charge.
Keep these circuits simple. Complicated circuits break and are tough to troubleshoot. I have repaired many protection circuits, many overly complicated. The amps were fine (of course).
-Chris
anatech said:
No, the actual amplifier sees a normal ground and load. The sense resistor is placed *between* the supply ground and the amplifier ground. (It may be smaller than 0.1 also)
All devices have a seldom appreciated non repetitive maximum rating well in excess of the steady state one. Electromechanical devices are even more resilient in this aspect.
A power supply that cycles with a low period (more than 1 second) and senses and shuts down in the milliseconds range qualifies.
Sometimes it is not possible to have both simplicity and performance. Complexity designed with a purpose and well done need not be unreliable. Take a look just inside your PC supply, scores of components, an horribly delicate balance on the brink of disaster yet mass produced under U$ 30 retail and reliable to boot.
Rodolfo
anatech said:
The series element will not lower damping factor because it is in the return ground of the power supply. The actual ground (signal and power) is connected before the element. It will, however, increase rail sag, and if it is a common power supply, introduce intermodulation signal of both channels into each of their power rails - hopefully PSRR and CMRR can suppress this. HF components may have to be shunted power/signal ground connecton via capacitors in order to even make the amps work properly, this will seriously degrade HF signal sensing on the resistor. A current transformer would definitely work better regarding rail sag but you may have a problem with LF response, which needs to be catered for. Also, if capacitors are used to shunt HF, you again have reduced HF current sensitivity.
ilimzn said:
True, on the other hand as you note, PSRR should routinely handle this, we are talking about less than 1V peak, or milivolts if the sensing resistor is brought down to the order of tens of miliohms since we are amplifiying the signal later.
True again, but we are concerned with overload detection in the millisecond or sub-millisecond time frame, so HF desensitization should neither be an issue.
Rodolfo
mikeks said:
Unfortunately PVI gate drives make MOSFET switching virtually as slow as an EM relay....insufficient current to charge the MOSFETs input capacitance fast enough....See pg 8 here:
http://www.irf.com/technical-info/appnotes/an-1068.pdf
Therefore i reckon floating supplies are the optimal solution....
http://www.diyaudio.com/forums/showthread.php?postid=625816#post625816
Re: General protection mumblings
Hi Rodolfo,
Being in the middle of an amp design, and studying mikeks' article on subj in an earlier EW, I have a slightly different opinion on this. I don't think I want to protect the speakers, from the amp point of view. We have no control over whatever speakers people connect to their amps, and an one-covers-all system will surely severely limit the amp.
On the other hand, if the amp is protected, and never will produce high DC levels, or extended periods of full power AC, we may expect the speaker to cope with whatever this amp delivers, no?
So, I say, start with the amp, make sure it "behaves", and make sure that if an output component breaks, the damage is contained. These really are two requirements, and possibly need two separate systems.
Jan Didden
ingrast said:
Hi Rodolfo,
Being in the middle of an amp design, and studying mikeks' article on subj in an earlier EW, I have a slightly different opinion on this. I don't think I want to protect the speakers, from the amp point of view. We have no control over whatever speakers people connect to their amps, and an one-covers-all system will surely severely limit the amp.
On the other hand, if the amp is protected, and never will produce high DC levels, or extended periods of full power AC, we may expect the speaker to cope with whatever this amp delivers, no?
So, I say, start with the amp, make sure it "behaves", and make sure that if an output component breaks, the damage is contained. These really are two requirements, and possibly need two separate systems.
Jan Didden
Re: Re: General protection mumblings
Hi Jan, good to hear from you again:
I basically see no contradiction. I understand you delete the speaker ratings from the design specifications and leave only the amplifier (basically power output stage) protection and failure containment.
Unless I am missing something, the end result from my proposed scheme is the same.
Rodolfo
janneman said:
Hi Jan, good to hear from you again:
I basically see no contradiction. I understand you delete the speaker ratings from the design specifications and leave only the amplifier (basically power output stage) protection and failure containment.
Unless I am missing something, the end result from my proposed scheme is the same.
Rodolfo
Re: Re: Re: General protection mumblings
Not really....
Many amps combine SOA protection with the DC protection system...using the relay to disconnect the speakers in the event of SOA violation, or DC overload...
The difficulty here is that a relay cannot hope to open fast enough to give foolproof SOA protection...In fact, this sort of arrangement is largely cosmetic...!!!
Therefore, use purely electronic, self-resetting SOA clamp-protection...entirely independent of other protection systems...
There is no excuse for slack design in this regard....
ingrast said:
Not really....
Many amps combine SOA protection with the DC protection system...using the relay to disconnect the speakers in the event of SOA violation, or DC overload...
The difficulty here is that a relay cannot hope to open fast enough to give foolproof SOA protection...In fact, this sort of arrangement is largely cosmetic...!!!
Therefore, use purely electronic, self-resetting SOA clamp-protection...entirely independent of other protection systems...
There is no excuse for slack design in this regard....
Hi Jan,
That's been my point all along. I the user abuses the amp, tough for him. If the amp becomes defective it should not be allowed to damage anything else. Two detection systems is reasonable and my thought as well.
-Chris
That's been my point all along. I the user abuses the amp, tough for him. If the amp becomes defective it should not be allowed to damage anything else. Two detection systems is reasonable and my thought as well.
-Chris
Twony pages of discussion and still any concrete schematic. Your productivity isn't miracle, guys 🙂 .
- Status
- Not open for further replies.