Hello folks,
I ain't no reliability EE but I do remember having requirements for HALT/HASS testing on some rather big systems. It was a design requirement in most part to satisfy ourselves and the likes of customers, i.e. EMC or IBM. How do you know a systems MTBF? do calculations per mil specs or do destructive/stress testing. How else would you know where the weak links are?
Thermotron environment chambers, accelerometers = shake, bake, rattle and roll.
God, you guys are basically talking audio electronics here, how many mfg's in this field, actually do complete stress testing? If one audio amp blows, in the field, you ain't taking away $M's of daily business for the customer and a load full of irate customers nationwide.
I use crowbars on supply rails where they really count and I am usually the only customer. I do not think I have ever seen one audio product incorporate a supply crowbar. Only ones, I see commonly, are on older HP test equipment. The cost of a zener/scr far out way the cost of all those expensive parts down stream. A single point of failure causing huge amount of $'s and time to repair.
I did see that Doug Self did incorporate a supply shutdown on the Elektor NE5532 based PA, so that if one rail failed, it would shutdown the complemenary rail = good idea, thanks for the great idea, yet simple ckt, but it once again does fall short, as there was no OVP crowbar, oh well, take yer chances I guess, NE5532's are cheap and I assume purposely socketed.
I guess the design requirement for the most part in audio gear, was/is if the supply goes OV, who cares what is blown down stream, obviously of no concern to the designer's or organization, poor customer is SOL for the cost of a $1 or so in extra parts.
Cheers Rick
I ain't no reliability EE but I do remember having requirements for HALT/HASS testing on some rather big systems. It was a design requirement in most part to satisfy ourselves and the likes of customers, i.e. EMC or IBM. How do you know a systems MTBF? do calculations per mil specs or do destructive/stress testing. How else would you know where the weak links are?
Thermotron environment chambers, accelerometers = shake, bake, rattle and roll.
God, you guys are basically talking audio electronics here, how many mfg's in this field, actually do complete stress testing? If one audio amp blows, in the field, you ain't taking away $M's of daily business for the customer and a load full of irate customers nationwide.
I use crowbars on supply rails where they really count and I am usually the only customer. I do not think I have ever seen one audio product incorporate a supply crowbar. Only ones, I see commonly, are on older HP test equipment. The cost of a zener/scr far out way the cost of all those expensive parts down stream. A single point of failure causing huge amount of $'s and time to repair.
I did see that Doug Self did incorporate a supply shutdown on the Elektor NE5532 based PA, so that if one rail failed, it would shutdown the complemenary rail = good idea, thanks for the great idea, yet simple ckt, but it once again does fall short, as there was no OVP crowbar, oh well, take yer chances I guess, NE5532's are cheap and I assume purposely socketed.
I guess the design requirement for the most part in audio gear, was/is if the supply goes OV, who cares what is blown down stream, obviously of no concern to the designer's or organization, poor customer is SOL for the cost of a $1 or so in extra parts.
Cheers Rick
Great stuff -them MOSFETS
These are my bookmarked links for when I want to build some....
The fundamentals -
ESP - MOSFET Solid State Relays
A DC Fault Protection Circuit for Audio Amplifiers
Many DIYA successful implementations -
http://www.diyaudio.com/forums/solid-state/237730-solid-state-speaker-protection-system-interest-check.html
Lots of info , almost too many component choices (SMD - here I come
)OS
"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."
If the speaker is disconnected ( fuse open) no problem.
It's like marriage. If your wife is yelling at you from the kitchen, just get up and close the door. Problem solved
If the speaker is disconnected ( fuse open) no problem.
It's like marriage. If your wife is yelling at you from the kitchen, just get up and close the door. Problem solved
Fair enough, if you design a bridge to hold 10 tons, it's a good test to load it until it breaks to see that it holds at least 10 tons.
But the discussion was about DESIGNING by destructive testing, which IS pretty stupid. You wouldn't design a succession of bridges and load them till destruction until you are satisfied.
jan
Last edited:
Except that many amplifiers object to prolonged hard clipping. You would need an extra protection circuit to detect this and cut the signal to the amplifier.
I think I will build the Land Rover amp and give it to the lad who's Rotel I keep fixing . He is a musician . He can have a new one to brake . I think a radical up date of my beloved Quad 303 wouldn't go amiss . The basics are good . Now all it needs to do is drive 1 ohm . Why not give it 24 x 2N3055 ? I bet it will sound better than many reading this would imagine . Capacitor coupled and tested to destruction . Celestion had capacitors in the DL speakers . They worked very well . I suspect they had a nicer quality of bass due to it . They were only 500 uF !!!! We seldom had to fix any ( never ? ) . Mission all the time . Sorry to repeat myself . My new amp would have a lot of this in it in the link below .
MJR7-Mk5 Mosfet Power Amplifier
MJR7-Mk5 Mosfet Power Amplifier
We test the components instead . Less mass also to handle . We monkey about with aircraft also in the same way . The number of lab reports I did on proof stress and plastic deformation . Andrew is right . An amplifier is cheap enough to call a component .
Modern buildings have acceptable amounts of cracking , bridges also . To the untrained eye it would look bad . It is how far they have traveled and starting from where . Concrete is poor in torsion , strong in compression . The reinforcing bars often mark the stop point . If so 500 years might not see a change (doubtfully ) . The big deal is the cracks appear soon after the completion of the building . If they didn't appear that would be a greater mystery . The Tay Bridge disaster caused a greater understanding of such problems . Pig iron substituted for steel I think was the likely cause . It looks the same when painted .
There were some amps. For example, the Studer A65 / Revox A740 :
http://www.pure-hifi.info/pdf/Le magazine audiophile -Analyse du Revox A740.pdf
I wish I could find an advert I loved from the past . A man with a pipe and a bucket of water and smoke coming form his speakers " Should have bought a Bryan amplifier "
Bryan never said what the fantastic protection circuit was . A capacitor .
The Bryan was the Tobey and Dinsdale design . Leak also was similar . I think someone told me Mr Leak was overheard speaking to either Mr T or D and saying . Look I am a very rich man and if this get to court you will not win because I am a very rich man . More to the point T&D were building on the work of Mr H C Lin without reference also to that . To me publishing in Wireless World is as good as giving it away . I have doubts it was Mr T or D . perhaps they just wanted to be said to be the originators ? Mullard also for the valve amps ? No , they were happy to sell the valves .
Alan Tisdale told me Harold was crying on the shop floor on the Monday after he sold the company for £ 3 000 000 ( price of 1000 houses then or 10 000 BSA A65's Thunderbolts ) . The staff were away at a show when the Rank people descended on him . My feeling is he should have bought it back a few years later when Rank had ruined it . People seldom speak of Leak . Out of the ashes came Rotel UK .
H.J.Leak and Co. Ltd.
http://44bx.com/leak/1968.html
Bryan never said what the fantastic protection circuit was . A capacitor .
The Bryan was the Tobey and Dinsdale design . Leak also was similar . I think someone told me Mr Leak was overheard speaking to either Mr T or D and saying . Look I am a very rich man and if this get to court you will not win because I am a very rich man . More to the point T&D were building on the work of Mr H C Lin without reference also to that . To me publishing in Wireless World is as good as giving it away . I have doubts it was Mr T or D . perhaps they just wanted to be said to be the originators ? Mullard also for the valve amps ? No , they were happy to sell the valves .
Alan Tisdale told me Harold was crying on the shop floor on the Monday after he sold the company for £ 3 000 000 ( price of 1000 houses then or 10 000 BSA A65's Thunderbolts ) . The staff were away at a show when the Rank people descended on him . My feeling is he should have bought it back a few years later when Rank had ruined it . People seldom speak of Leak . Out of the ashes came Rotel UK .
H.J.Leak and Co. Ltd.
http://44bx.com/leak/1968.html
Last edited:
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.
Of course, you don't test to destruction on the final form amp, rather on the working model.
I have seen products delivered from famous name manufacturers which performed differently from their spec. Later on, it turned out that was a funny batch, and indeed, subsequent piceces were up to scratch.
Some years ago, I received a batch of famous name op amps. In general, they worked fine, but consistently demonstrated a 0.6V offset. Now, of course you can compensate for it, but you shouldn't have to, and if you relied on the spec sheet only, you'd be in trouble to add offset compensation.
I double checked myself by sending the circuit schematic to a friend from UK, who also found the odd offset. Then I got in touch with the manufacturer and informed them of it, stating that a simple 1N4148 diode (anode to -Vcc, cathode to output pin) solved the problem. I got no answer. But later versions of the op amp did not have the said offset.
My point is, you cannot rely of the spec sheet absolutely, it's too risky. That's why you need to push it to burnout to verify you don't happen to have a funny batch, or some such. Obviously, this applies to power transistors only, you don't need to burn out op amps and such, just drive them to near clipping.
And, these days, you always need to verify your devices, given just how many duds are offered from the Far East. I've seen far too many times devices rated at 200+V give up the ghost at 40 or 50V simply because they were duds.
Rick, I make my products to last at least 20 years of 24/7 use. I know this is no longer expected, but I am 60 years old and have gone to the old school, which taught me to build it like a rock.
I can't help it, that's the way I was shaped in my youth, that's the only way I know how to do it. In the last 12 years, my current line of power line filters (currently called all but their real name), NOT ONE has failed anywhere in the 28 countries I have sold them to. Not one.
I intend to keep it that way, that's a statistic fact I am very proud of and which I have worked hard for.
No Jan, YOU turned it to that.
I never said I design by destruction, I simply said, and I repeat now, after all the testing is done, the fate of the first prototype is to get blown away by being pushed to the limits of its protection circuits, and then beyond. I have to know its absolute limts before destruction in order to say either this is fine, or I need to change this, just in case. Not rely on the data sheet, or maths, but on a living model, configured to work as it I planned it to be.
You might say I tend to overdo it with this - you might be right, I don't think so, but I believe I'm simply being cautious. Or, it might be that in these days of hit-and-run products, not many bother doing all that in their mad rush to into the market.
Actually, I don't do it much any more, after so many years of using the same power transistors always from the same manufacturers, I pretty well know how they work and what can they take. And I do tend to overdo the heat sinks (by commercial standards, that is), but then, I also tend to overdo the whole power supply block (transformer-rectifier-capacitor). With this in mind, most of what I do has inside it all the ingredients of possible self-destruction, so caution is called for.
Hi dvv & others,
Wonderful that you have a high degree of concern towards a quality product. When you wear all the hats then you can make this call or requirement. Not so, when the bean counters get involved.
Must not confuse initial product verification vs continual quality control = a different set of requirements.
At Motorola back in 1980's, the high end radio (IMR) got a 12 hour EIA temperature cycle with a selected set of important parametric tests. Yes, a Thermotron full of radio's, up to 144, at one time, I helped build/maintain the setup. Of course there were failures, that might not have ever shown up in the field or many years later. This ended, as it was just too costly, time consuming vs the perceived gains, so then they backed it down to lot sample testing, as a viable alternative. I think incoming component QA went away at or about the same time. I remember those Tek curve tracers, 575? IIRC
Cheers
Rick
Wonderful that you have a high degree of concern towards a quality product. When you wear all the hats then you can make this call or requirement. Not so, when the bean counters get involved.
Must not confuse initial product verification vs continual quality control = a different set of requirements.
At Motorola back in 1980's, the high end radio (IMR) got a 12 hour EIA temperature cycle with a selected set of important parametric tests. Yes, a Thermotron full of radio's, up to 144, at one time, I helped build/maintain the setup. Of course there were failures, that might not have ever shown up in the field or many years later. This ended, as it was just too costly, time consuming vs the perceived gains, so then they backed it down to lot sample testing, as a viable alternative. I think incoming component QA went away at or about the same time. I remember those Tek curve tracers, 575? IIRC
Cheers
Rick
You don't need to test to destruction.
Semi manufacturers test their products ('qualification') and publish data sheets. Some of this testing at component level may be to destruction. There is a whole science around this aspect in the industry with some very solid mathematics behind how the results are interpreted.
The only thing you need to do as a product designer is work within the data sheet specs, and design conservatively. For passive components, there is a similar work methodology to qualify components.
Outside of cheap consumer products, the industry does a remarkable job in my view and quality is very high, with failures in the ppm level, with many product categories in the low ppb (25 to 50 ppb). Most problems are because of over stressed components coming from marginal or bad designs.
I had a big customer in J that wanted to use some 100 V power mosfets in an application that would expose them to 110V for a consumer application. I refused. He asked me to select I told him the yield would be bad and still refused. Eventually he went away and used a competitive product - hopefully it was rated for the job. Just one small example of bad design practice and there are many others . . .
Semi manufacturers test their products ('qualification') and publish data sheets. Some of this testing at component level may be to destruction. There is a whole science around this aspect in the industry with some very solid mathematics behind how the results are interpreted.
The only thing you need to do as a product designer is work within the data sheet specs, and design conservatively. For passive components, there is a similar work methodology to qualify components.
Outside of cheap consumer products, the industry does a remarkable job in my view and quality is very high, with failures in the ppm level, with many product categories in the low ppb (25 to 50 ppb). Most problems are because of over stressed components coming from marginal or bad designs.
I had a big customer in J that wanted to use some 100 V power mosfets in an application that would expose them to 110V for a consumer application. I refused. He asked me to select I told him the yield would be bad and still refused. Eventually he went away and used a competitive product - hopefully it was rated for the job. Just one small example of bad design practice and there are many others . . .
God, no bean counters for me, thank you, Rick. I am fiercly independent all the way, my company, my trade mark, my decisions - period. As for how I do things, ask my employees and suppliers, when was payment and paycheques late by even one day. Ask my bank if I ever took out a loan.
Actually, the simple truth is that I am a bean counter by vocation, as I have a degree in economics. However, my logic is simple - make it really good, give it a fair price and people will talk, generating trade. Currently, you cannot find me on the Internet, the site is off. True, a new one is being made, but believe it or not, my past trade is generating nice jobs for me all the time, with just short pauses in between. My No.1 market is an 11 year old battle between the Netherlands and Germany, trading places every year.
Obviously, I am small time, I do only hand crafted products, and I only have two products, power line filters and headphone amps. And since quality by defalt assumes a price, I was never likely to get rich from it. But, for what it's worth, I love my job. I love building things.
And I'm no newbe, the first filter was made way back in 1974, a time when most making them today didn't even exist as companies. Of course, today's versions only remotely resemble the Grandpa of those days, but the underlying principle is the same. When I hear a better one, I might change.
With all this in mind, I came to the conclusion that while not strictly necessary, burnout limit testing is a very good idea, if your entire manufacturing philosophy is based on security and longevity. I would take any one of my products failing in less than 10 years at the very least, normally 15 years and more, as my own failing. Having burnt a few, I have learnt things which helped me improve the product to the point when the new version survives the conditions the old version didn't.
Now, Jan for example feels differently, as is his undisputed right. I do not claim that my way is the only, or even the best way, but it sure works for me. Much depends on one's general views regarding build quality, and ultimately, even the best device can break down, so all I am really doing is trying to minimize the chances of it breaking down. Doing what I can. For whatever it's worth, but it means a lot to me personally. Quite simply, I want to be proud of my products.
Naim did test to destruction . I am told a standard test done for days had the emitter resistors glowing in the dark . These replaced on dispatch . I never had an NAP 250 go wrong . Looking back on it Naim and others fooled us . They hinted at things that no customer required . When building gradient coils I used the 3 phase powered Amcron . What we did to that amp I don't want to say . We used every once of current it had into far less than an ohm . Apart from the cable that went in it looked smaller than some Audiophile designs . The fan kicked in if required . I can borrow one any time I like ( 2 phase ) . John who owns it said it is not the best sounding . John usually is reluctant to say that so one must say it is a Tractor and not a Landrover . His best suggestion was to base my Landrover on C-Audio of old . I had that idea also . John said at gigs he would repair them and think nothing of it whilst doing the mixing , now with class D they carry spare amps . The amp would never have multiply failure ( VAS etc ) . I heard an amp based on C-Audio from his boss ( David Mate SSL ) that Eric Braithwait reviewed . It was shunning . Apparently getting in back was a slow job . David said something I loved . " Seeing as we are going to use feedback who cares how good the pre-feedback linearity is " . He went on to say . The change in gain over minute signal levels matters most and the ability to use feedback . His strange and though provoking statement was industial triodes used in switching looked ideal . He made a synthesis of FET and bipolar to do it . I said to John did he use a the daft version of a complementary feedback pair ( FET in bipole out ) . John smiled .
At my former job, we did destructive, and Has/Halt test as well, both electrical and mechanical (vibration and shock), and as the equipment was used outdoors we also conducted IP and saline testing. For amplifiers in indoor environment not all those tests are needed, but not looking from most angles will over time cause you trouble. I for one would love to avoid the smallish ill smelling fumes of output stage burnout, looking to get the grasp on SOA protection
I notice those who design as if you would have problems seldom do . My old boss was a very highly trained mechanical engineer . He had 7 years university before even starting in Vienna ( 1937 ) . He left as he was a communist . I asked him what was Hitler like " better looking than you might imagine " was the unexpected reply . Ironically he was captured by the Soviets and put in one of their concentration camps to help out Hitler ( German paperwork sent to the soviets during their brief pact ) . I went to him as an electrical engineer with a hobby in electronics good enough to do the job . I have an electronics diploma also , not time served at the time ( 1974 ) .The deal was I would teach him electrical engineering and he teach me mechanical . Nothing much happened except I had it drummed into me how true destruction testing is done . Most of the useful data is at 150 % expected load as a generalization . Sometimes we do better . One simple test was nitric acid . It was the only way of saying to the plating company the job was not to standard . 40 microns gold will not be destroyed by nitric acid . Do it wrong and it will . It is surprising how many suppliers agree a contract they can not deliver . These connectors take vast current and do not degrade with time . They need almost zero insertion force . All the gold did was keep it that way . The gold held up production for years to get it right . The company had made a disaster over a ferromagnetic particle clutch . Advancing my boss to the connectors division was basically giving him the sack for protesting too much over the clutch that was bought by Toyota in the end . The company one the UK 's largest suppliers to the motor industry is now no more . The connectors live on .
The best one I heard from Terry O'Sullivan . He was technical sales rep for Wolf power tools ( same area of London as Leak ) . Their hammer drills were splitting . None of this was happening in the rigors of testing . Finally after all the technical staff had failed Terry was sent to British Railways Swindon to see what the problem was . Terry already thought it curious that they were the ones with most problems . After a long and reasoned chat with the service manager Terry asked if he could see the job that caused the problems . Along comes the expected mountain of a man who one treats with respect at all times . Terry asked timidly " could you show me how you use the drill " . You might have already guessed it . The drill was switched on and in the other hand a hammer ! Terry obviously was communicating by looks alone his thoughts . The big man answers as he hits the drill " well mate , it is a hammer drill " . These things are beyond destruction testing .
The best one I heard from Terry O'Sullivan . He was technical sales rep for Wolf power tools ( same area of London as Leak ) . Their hammer drills were splitting . None of this was happening in the rigors of testing . Finally after all the technical staff had failed Terry was sent to British Railways Swindon to see what the problem was . Terry already thought it curious that they were the ones with most problems . After a long and reasoned chat with the service manager Terry asked if he could see the job that caused the problems . Along comes the expected mountain of a man who one treats with respect at all times . Terry asked timidly " could you show me how you use the drill " . You might have already guessed it . The drill was switched on and in the other hand a hammer ! Terry obviously was communicating by looks alone his thoughts . The big man answers as he hits the drill " well mate , it is a hammer drill " . These things are beyond destruction testing .
Last edited:
Defective SOA Protection
Douglas,
On page 311 you give a schematic of an amplifier with Fairchild's single-slope SOA protection.
Regrettably, with the component values given, the protection circuit offers virtually no protection as its locus intrudes heavily into the underated output transistor's power curve as shown in the attachment below.
Douglas,
On page 311 you give a schematic of an amplifier with Fairchild's single-slope SOA protection.
Regrettably, with the component values given, the protection circuit offers virtually no protection as its locus intrudes heavily into the underated output transistor's power curve as shown in the attachment below.
Michael . So interesting you say this . I had an application where protection was important ( industrial ) . In the end I added a under run CCS power supply . Simple and 100 % effective . Then I remembered a critical review of the Naim NAP 250 saying the famed power supply was nothing more than a constant current source . Correct and not correct . I find it a mystery why conventional current limiting is used . I doubt it works to anyone's benefit ? The type of CCS I refer to is one that will provide up to that current ( R load requiring far less ) . They are lovely in that they come on gently in most cases .
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.