Hi Jan. Good to hear from you; hope you're OK.
Ahh, could be better. But could also be worse. A lot.
Looking forward to meet Bob next month - maybe I can explain your SOA thing ;-)
jan
Last edited:
Do you seriously think a fuse will at all times manage to protect the speakers?
Do you equally seriously believe that many people just love changing burnt out transistors? Or even know how?
Isn't it wiser to test your amp first right to the point of destruction, and when you have learnt where it is, THEN design a protection circuit which will allow the transistors to get near the point of destruction, but not quite up to it? Remember, you can use a built-in delay which could well take you to the 50 mS line on the SOAR curve (though I personally prefer the 100 mS line, which I run with a 50 mS delay).
Do you equally seriously believe that many people just love changing burnt out transistors? Or even know how?
Isn't it wiser to test your amp first right to the point of destruction, and when you have learnt where it is, THEN design a protection circuit which will allow the transistors to get near the point of destruction, but not quite up to it? Remember, you can use a built-in delay which could well take you to the 50 mS line on the SOAR curve (though I personally prefer the 100 mS line, which I run with a 50 mS delay).
Been there, done that. Around 1984 IIRC.
Actually it was an analog computer, sort of, calculating instantaneous dissipation and comparing it to a model of the SOA. Worked well, but was a bit complex.
Audio Amateur published in their magazine, then later in their Power Amp book.
Later I reverted to a simpler design similarly to Mike's, although he maintains mine doesn't work. Which unfortunately I didn't know when I build it into my amp so it is just working away merrily. ;-)
Some info here http://www.linearaudio.nl/linearaudio.nl/index.php/my-projects/safe-operating-area-calculations
How about you? You published your protection design yet?
jan
Actually it was an analog computer, sort of, calculating instantaneous dissipation and comparing it to a model of the SOA. Worked well, but was a bit complex.
Audio Amateur published in their magazine, then later in their Power Amp book.
Later I reverted to a simpler design similarly to Mike's, although he maintains mine doesn't work. Which unfortunately I didn't know when I build it into my amp so it is just working away merrily. ;-)
Some info here http://www.linearaudio.nl/linearaudio.nl/index.php/my-projects/safe-operating-area-calculations
How about you? You published your protection design yet?
jan
Last edited:
Yes, the safety of the loudspeakers should always be the first priority.
A crowbar is fine, and it need not destroy an amplifier that is not already destroyed.
Cheers,
Bob
That reminds me..... besides spice models -- might be nice to do and show some tests besides mfr data -- like via a curve tracer on as many of the recommended transistors as possible.
-RNM
Actually that seems a pretty stupid way to design anything....
jan
Thank you for the compliment. My experience tells me it is not so, not if you want to push your protection circuits to the real world brink, but not over that.
I will agree that you can calculate it fairly near the real absolute limit, but pushing it on basis of calculations alone may cause you to cross the thin, red line without wanting to.
I know of instances when sticking to Data Sheets caused some burnt out power stages, because the manufacturer was very liberal in their specs. And it can be just a touch, for example, Data Sheet says you can use a say 3.3k resistor, but if you want to avoid some blue smoke, make that 3.3k a 3.6k. Very often, 3.3k should be 3.9k.
The trouble with protection circuits is that ultimately, if you want to be really sure, you have to test them to the limit. And sometimes, that limit is actually lower than suggested by Data Sheets. So, I do not see risking a pair of trannies to the point of destruction as being "stupid", not even a little bit. It's the only way I can think of which takes into account the true operating condition of the trannies, especially the thermal derating factor.
With your method you cannot test for minimum devices (unless you select a minimum device from a batch which is not realistic). This goes for many other parameters in your circuit/system - you have no way to insure you have a typical or minimum system.
Calculations allow you to design for minimum system/ maximum stress and ensures that the protection comes in before the device and your amp explodes.
I am still of the opinion that your method is not practical and does not ensure a good design.
When you say: "The trouble with protection circuits is that ultimately, if you want to be really sure, you have to test them to the limit." That's nonsense. Why do I need to test them to the limit to be absolutely sure? What does that bring me? Only one more data point from a very specific unit? How does that relate to the other 1000 units I am going to build? Precisely because data sheets show a variation, your only sure way is to design for minimum/maximum values and that cannot be done by blowing up a random unit.
jan
Calculations allow you to design for minimum system/ maximum stress and ensures that the protection comes in before the device and your amp explodes.
I am still of the opinion that your method is not practical and does not ensure a good design.
When you say: "The trouble with protection circuits is that ultimately, if you want to be really sure, you have to test them to the limit." That's nonsense. Why do I need to test them to the limit to be absolutely sure? What does that bring me? Only one more data point from a very specific unit? How does that relate to the other 1000 units I am going to build? Precisely because data sheets show a variation, your only sure way is to design for minimum/maximum values and that cannot be done by blowing up a random unit.
jan
Last edited:
A device may be good for 50W at 10V Vce but may only be good for 10W at 40V Vce. So it is a dynamically varying parameter and you can't just say double the devices. You need to look at the SOA really. Mike K has written an interesting article about it, you should read it. Or see my linearaudio.nl website under projects.
jan
In times gone by, we used to build and when it fell down, we knew the factor of safety was not good enough. This method proved to be very wasteful of resources.
Civilisation moved on to try to model the built world and used that model to guide the adequate factor of safety version that we wanted to build.
We MUST test to destruction to prove that the model closely represents the built world and what the factor of safety is.
The Romans built some structures that in hindsight had a very adequate factor of safety. As a result they have left us some wonderful sights of ~ 2000years old.
If we were building a similar structure, what would we adopt as a reasonable life span?
Would we test our model before building? Would we be confident that it would safely reach it's intended lifespan?
Moving to semiconductors: How do the manufacturers generate the data. I guess they test to destruction.
Testing to destruction is a fair method of determining that our models are fairly accurate.
Civilisation moved on to try to model the built world and used that model to guide the adequate factor of safety version that we wanted to build.
We MUST test to destruction to prove that the model closely represents the built world and what the factor of safety is.
The Romans built some structures that in hindsight had a very adequate factor of safety. As a result they have left us some wonderful sights of ~ 2000years old.
If we were building a similar structure, what would we adopt as a reasonable life span?
Would we test our model before building? Would we be confident that it would safely reach it's intended lifespan?
Moving to semiconductors: How do the manufacturers generate the data. I guess they test to destruction.
Testing to destruction is a fair method of determining that our models are fairly accurate.
Last edited:
Hi Bob . It's a bit like making an off road vehicle . The solutions are from the farm and the truck . Most people who think they know these things haven't spent years under Landrovers ( I had no choice , they need care and servicing , one has done 320 000 miles without one problem except a gearbox shaft which was replaced as a precaution at 180 000 ) . The Landrover concepts are so awful as to make it a non starter in engineering terms , as an admirer I still think that every time I look at one . We know the reality , it's called Range Rover . One of the few car companies making profit . I have a hunch if with great marketing I made the amplifier version of this many would become converts . All the Ferrari stuff would be for the birds as we used to say . The weird bit is the new Range Rover Sport is almost a decent road car . I think a French fashion magazine said the only vehicle to be seen in , others are bad taste .
Andrew T . George Stephenson couldn't read nor write like myself . He based his ideas on Roman engineering . His bridges still work for locomotives 5 times larger than he knew . Andrew you are right about doing the job correctly and testing to destruction . Robert his son had a proper education . He was a very good friend of I K Brunell . That shows engineers can get on . Robert and Isambard had every reason not to be friends . Robert often wrote letters when he knew his friend was in trouble ( 80 letters found I think ) .
Last edited:
The true and only acid test for any circuit is to drive it beyond its nominal requirements until it fails. If it fails at say 30% above nominal, we can say we have a fairly good ciruit, working with a solid margin of error.
I would like to meet a man who has always designed by maths only and has NEVER had to modify any one of his circuits to a value not available by calculations.
The need to test under actual operating conditions is further promoted by the fact that some manufacturers, now I refer specifically to Motorola/ON Semi, consistently lie about their products, fortunately for me, in their case in the right direction. I have yet to encounter a transistor made by them which did not exceed the nominal spec, if only by little. But I have never ran into any of their transistors which failed at nominal operating conditions.
Assuming only of course that it IS really one of theirs, not a Far East rip-off.
First it was just me, and now Andrew and Nigel ...
Seems a lot of peole are doing stupid things.
BTW, for the record, though you know it already Nige, I am not old, but ancient school, as I believe a safety margin of at least 50% to be the norm, and often strive to do better than the norm.
Perhaps it's those 3 years in UK schools which did me in ...
Seems a lot of peole are doing stupid things.
BTW, for the record, though you know it already Nige, I am not old, but ancient school, as I believe a safety margin of at least 50% to be the norm, and often strive to do better than the norm.
Perhaps it's those 3 years in UK schools which did me in ...
If you have yet to establish the matrials, design and models of your structure or process, then obviously you need to explore every unknown, relevant avenue to establish them if you expect a succesful outcome. That will certainly mean destructive testing. If we look in museums, memorials and documents preserving monumental works of the past, you can often see twisted beams, girders, riveting, welds, hawsers etc. tested to destruction but finished structures? Nope!
We place a few loads here and there, drive some huge vehicles etc. across and monitor for years to come, predicting with improving computer models as technology marches on.
Certainly, as a basis of establishing a design or a component's integrity, we have to go down to the fundamentals of proof testing. However, I think Jan's message is relavant to production processes. The idea there is to establish and control the process such that final testing, particularly destructive testing, becomes redundant. With a true quality system in operation, paramaters are more or less locked in to the intrinsic behaviour of the devices, as we see in the differences between BJTs and relatively unpredictable JFETs. BJTs, for instance, are now extremely well controlled, consistent etc. and i doubt you'll see much destructive testing in transistor production plants. I imagine there are also related properties that correlate with some destructive tests , making them unnecessary, too.
Hello Bob
Have to disagree with you here. I don't know if you have any experience of using crowbars on amplifiers, but I do, and it is all bad. It very often converts what might be a minor fault in an early stage of the amplifier into total destruction. You have to dump maybe 100,000 uF of charged capacitance into a triac or whatever and unless it is enormous it will explode before the DC rail fuses can react. This current has to go through the output devices and they are likely to be destroyed if they are not dead already. PCB tracks can disappear in a blue flash. All in all it's a brutal business.
I have always used output relays for my commercial designs and there have never been any problems there. Since you almost certainly need a mute relay to stop thumps and bumps at turn-on and turn-off, there is no extra cost.
Last edited:
Relays are also problematic in decent sized amplifiers. I know - I took out a B&W bass unit because the output stage failed, the relay tried to open, the contacts arced and then fused, putting 80 V directly across the speaker. Cost me circa 300 GBP to fix the speaker. http://www.diyaudio.com/forums/solid-state/191449-output-relays.html
For the triac trick, try a FB fuse in the speaker line before the triac. Take feedback after the fuse. Job done.
I've moved on the mosfets now. Might be a bit expensive for mass market amps, but works a treat and can cleanly switch inductive loads up to 500mJ all day - a relay can do that once and will probably need replacing. Here's my write up http://hifisonix.com/wordpress/wp-content/uploads/2012/08/Speaker-Relay-V1.03.pdf
For the triac trick, try a FB fuse in the speaker line before the triac. Take feedback after the fuse. Job done.
I've moved on the mosfets now. Might be a bit expensive for mass market amps, but works a treat and can cleanly switch inductive loads up to 500mJ all day - a relay can do that once and will probably need replacing. Here's my write up http://hifisonix.com/wordpress/wp-content/uploads/2012/08/Speaker-Relay-V1.03.pdf
Last edited:
This certainly can happen if the relay is not up to the job. Selecting a relay that is both capable and cost-effective is one of the challenges of practical amplifier design.
But it would need to be a pretty big FB fuse to avoid it blowing on bass transients, and if it did blow, the amplifier would be deprived of feedback and go mad, quite possibly hurting itself.
I used 16A 250 VAC relays. Problem is very few relays handle DC faults like the one I had - auto relays done cut it either.
The MOSFETs work really well. Note in the 2nd Lin I posted, the ground line is switched (as done on my nx-Amp). I have also switched the speaker hot line - that's done on my e-Amp.
The MOSFETs work really well. Note in the 2nd Lin I posted, the ground line is switched (as done on my nx-Amp). I have also switched the speaker hot line - that's done on my e-Amp.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.