How fast does offset protection have to be

[Mooly]

Wondered if anyone has any info on how fast speaker protection has to be in the event of a fault. If you were to apply say 40 volts.dc across a typical speaker how long ( we are talking miiliseconds ) could it withstand this without damage.
I ask this because I am trying to develop a different type of offset protection that eliminates any RC integrating network time constant.
Karl

[Kostya-M]

Protection circuit can’t confuse full amplitude signal with lowest frequency. If we accept 20Hz, time must be greater then 25ms. One speaker passed this test, but another can get mechanical damage.

[Greg Erskine]

Hi Mooly,

Have you read Rod Elliott's Project 33:

http://sound.westhost.com/project33.htm

regards

[unclejed613]

the general rule of thumb is that the DC detector has a cutoff frequency of 5hz. the use of RC time constants and some threshold (such as 500mV) does not necessarily mean that it will take 200mS for a relay to shut off if 40V DC appears at the output. it might take 200mS for 1V DC to shut the relay off, but with 40V DC, the rate of change through the time constant is 40 times greater, so the relay will shut off in 5mS.

[Mooly]

Hi Kostya-M,

Originally posted by Kostya-M
Protection circuit can’t confuse full amplitude signal with lowest frequency. If we accept 20Hz, time must be greater then 25ms. One speaker passed this test, but another can get mechanical damage.

This is one of the classic problems with conventional offset protection, selecting a suitable time constant. I have been trying to develop a circuit that gets round this by comparing input/output to the power amp. There is no lower frequency limit and the trip time is very fast, 10's of microseconds for the detector, the relay is another matter.
I guess what I am really asking is if you could apply 40v D.C. across a typical speaker for say 1ms then 2ms and so on at what point does damage occur.
We all hope we never need to find out how good are protection is, and there are all these designs out there (commercial and D.I.Y.) but are any actually tested. Take a screwdriver, poke the output transistor, does the speaker survive. As far as I am aware you never see total trip times for these designs.
Thanks for your interest,
Karl

[Mooly]

Hi Greg,
I have not seen Rods article on this but will have a read later today.
Regards Karl

[Mooly]

Unclejed, I had'nt calculated the rate of the exponential voltage rise on the cap. My maths is useless I'm afraid, should have paid more attention at college. When you say 5ms to reach 1 volt we still have the relay drop out time to add to this. I have seen some figures quoted for this, about 2ms I think. I am trying to eliminate the 5ms part of the problem, and then to see if I can speed up the relay drop out time by applying a "short" via a MOSFET, across the coil. The instant the coil voltage drops out the MOSFET pulls in. I have not tested this part of the design yet, been trying to get the detector to work reliably.
There will probably be more circuitry in the protection than in the amp
Regards Karl
p.s. It was much easier with a coupling cap

[AndrewT]

Hi,
using 40Vdc as your worst case seems to indicate you're looking at 50W to 100W speakers.
The drivers inside these are likely to have less than 50mm voice coils, probably a mix between 16mm and 35mm diameter.

These will not last long for a prolonged DC offset. But I suspect no damage from tens of mS and maybe even from a very few hundred mS.

How about testing using Fuses as your timed pulse interrupter?

A 1A fuse from a charged 40Vdc supply to a driver VC. See what happens! Or start with a F200mA and work your way up.

[Kostya-M]

Hi, Mooly!
There are several aspects of your question.
1. If you want to select only amplifier damage. Then your differential method is better of all.
2. If you want to prevent speaker damage. Then you must know speaker power limit as function of frequency.
3. Speaker damage can be produced by input signal offset if amplifier has no low frequency limit.

[Nordic]

Do not forget to add the time the realy chosen, takes to switch... this varies greatly too.

Ag
Mooly I see now you saw that ...