Big crowbar module for <4VDC?
in [Design]
|
Prev: Ultra-simple discrete 10W audio power stage for guitar use - would this work?
Next: Mitashi 4 GB Mp4 Player - MW156
From: Jim Thompson on 18 Dec 2008 13:09 On Thu, 18 Dec 2008 09:53:45 -0800, Joerg <notthisjoergsch(a)removethispacbell.net> wrote: >Jim Thompson wrote: >> On Wed, 17 Dec 2008 15:44:48 -0800, Joerg >> <notthisjoergsch(a)removethispacbell.net> wrote: >> >>> Hello All, >>> >>> After modding a 5V linear to 3.3V I feel a bit iffy about the crowbar, >>> considering the Dollar amount that is going to be connected to it. The >>> original crowbar is partially under (!) a big electrolytic, a bear to >>> modify. Plus it probably isn't a good one to begin with, couldn't see a >>> TL431 in there. >>> >>> So, is there a big old fat crowbar module anywhere that can wrestle an >>> 18A linear supply? The usual OVP modules from Condor and others only >>> have a smallish TO220 SCR in there which would probably go kablouie >>> before it can protect anything. >> >> Second question. You _are_ aware that the crowbar goes on the _input_ >> side of the linear regulator, aren't you ?:-) >> > >Ideally yes. Reality, no. Most crowbars are across the output terminals. >They do their job because typically the supply is toast anyway if the >crowbar has to kick in. IOW one of the pass transistors has fused shut >and then it doesn't matter where the crowbar sits. Nonsense! Regulator loop fails, pass device can't pass enough current to blow fuse. > > >> Short series loop of input electrolytic and fuse and SCR... phzzzzt >> ;-) >> > >And that's the problem ... the SCR needs to be huge. Again nonsense! Sounds like you've never actually designed a crowbar-protected supply ?? I did my first 33 years ago, before simulation capability, worked perfectly ;-) >There is a really >fat one in the supply (>>TO220) but to get there would require major >surgery. Doesn't take a big "fat one". The clue to your learning process is I^2*t ...Jim Thompson -- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine Sometimes I even put it in the food
From: Joerg on 18 Dec 2008 14:16 Jim Thompson wrote: > On Thu, 18 Dec 2008 09:53:45 -0800, Joerg > <notthisjoergsch(a)removethispacbell.net> wrote: > >> Jim Thompson wrote: >>> On Wed, 17 Dec 2008 15:44:48 -0800, Joerg >>> <notthisjoergsch(a)removethispacbell.net> wrote: >>> >>>> Hello All, >>>> >>>> After modding a 5V linear to 3.3V I feel a bit iffy about the crowbar, >>>> considering the Dollar amount that is going to be connected to it. The >>>> original crowbar is partially under (!) a big electrolytic, a bear to >>>> modify. Plus it probably isn't a good one to begin with, couldn't see a >>>> TL431 in there. >>>> >>>> So, is there a big old fat crowbar module anywhere that can wrestle an >>>> 18A linear supply? The usual OVP modules from Condor and others only >>>> have a smallish TO220 SCR in there which would probably go kablouie >>>> before it can protect anything. >>> Second question. You _are_ aware that the crowbar goes on the _input_ >>> side of the linear regulator, aren't you ?:-) >>> >> Ideally yes. Reality, no. Most crowbars are across the output terminals. >> They do their job because typically the supply is toast anyway if the >> crowbar has to kick in. IOW one of the pass transistors has fused shut >> and then it doesn't matter where the crowbar sits. > > Nonsense! Regulator loop fails, pass device can't pass enough current > to blow fuse. > Not nonsense at all. Here's why: The resistance of the pass device, for whatever reason it is there, is desired. It makes sure the SCR doesn't turn into a plume of smoke. The pass devices will likely blow but who cares? Saving the $0.50/each pass devices is not the issue, saving a $10k load is the issue. Also, with OEM supplies you do not have access to the input side. Else the warranty is toast. Unless qties are huge and you can negotiate that "minor" inconvenience away. Not all issues are electronic ;-) Trying to short out a bank of serious electrolytics is IMHO not a smart thing to do. First, they guarantee max ESR but usually never a min ESR. Secondly, the trace resistance max can be calculated but not the min because PCB fabs usually do not guarantee a max copper weight. You've spec'd 2oz but what if it's a lot more? As an aside you'd also run the risk of exploding an electrolytic and if something bad results from that, oh boy, plaintiff's counsel would be all over you because I am sure they can source the same state-of-the-art I have (schematics of major supply mfgs) and use that as exhibits. Hint: There are emitter load share resistors and current sense resistors which further limit the damage. Puff of smoke versus major kablouie ;-) >> >>> Short series loop of input electrolytic and fuse and SCR... phzzzzt >>> ;-) >>> >> And that's the problem ... the SCR needs to be huge. > > Again nonsense! Sounds like you've never actually designed a > crowbar-protected supply ?? > Au contraire. And they all tested out fine. I would never, ever, consider placing a crowbar directly across a bank of serious input side electrolytics. That is IMHO a recipe for disaster. > I did my first 33 years ago, before simulation capability, worked > perfectly ;-) > >> There is a really >> fat one in the supply (>>TO220) but to get there would require major >> surgery. > > Doesn't take a big "fat one". It does. There is a good reason why they selected a much fatter one on this 18A supply. If I could get at it I'd be home. And if there ain't any COTS modules I'll have to. > > The clue to your learning process is I^2*t > I know about that. But business reality in systems architecture is often like this: There is no fuse before the regulator, only one before the transformer. Meaning the caps are free to dump their charge and good ones do so with gusto. You can place a fuse between the crowbar and the output of the supply but you must assume that it'll be an automotive rated slow fuse and that the pass devices have fully fused shut. IOW worst case, only load share and current sense resistors left in the game until the fuse decides to finally blow. I have here various schematics of OEM linears from reputable US manufacturers (unfortunately not at liberty to post on the web). All have the SCR connected to the output rails, just like I always did. Now all those manufacturers can't be wrong, can they? -- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
From: Jim Thompson on 18 Dec 2008 14:59 On Thu, 18 Dec 2008 11:16:38 -0800, Joerg <notthisjoergsch(a)removethispacbell.net> wrote: >Jim Thompson wrote: >> On Thu, 18 Dec 2008 09:53:45 -0800, Joerg >> <notthisjoergsch(a)removethispacbell.net> wrote: >> >>> Jim Thompson wrote: >>>> On Wed, 17 Dec 2008 15:44:48 -0800, Joerg >>>> <notthisjoergsch(a)removethispacbell.net> wrote: >>>> >>>>> Hello All, >>>>> >>>>> After modding a 5V linear to 3.3V I feel a bit iffy about the crowbar, >>>>> considering the Dollar amount that is going to be connected to it. The >>>>> original crowbar is partially under (!) a big electrolytic, a bear to >>>>> modify. Plus it probably isn't a good one to begin with, couldn't see a >>>>> TL431 in there. >>>>> >>>>> So, is there a big old fat crowbar module anywhere that can wrestle an >>>>> 18A linear supply? The usual OVP modules from Condor and others only >>>>> have a smallish TO220 SCR in there which would probably go kablouie >>>>> before it can protect anything. >>>> Second question. You _are_ aware that the crowbar goes on the _input_ >>>> side of the linear regulator, aren't you ?:-) >>>> >>> Ideally yes. Reality, no. Most crowbars are across the output terminals. >>> They do their job because typically the supply is toast anyway if the >>> crowbar has to kick in. IOW one of the pass transistors has fused shut >>> and then it doesn't matter where the crowbar sits. >> >> Nonsense! Regulator loop fails, pass device can't pass enough current >> to blow fuse. >> > >Not nonsense at all. Here's why: The resistance of the pass device, for >whatever reason it is there, is desired. It makes sure the SCR doesn't >turn into a plume of smoke. That statement proves you don't understand I^2*t >The pass devices will likely blow but who >cares? Saving the $0.50/each pass devices is not the issue, saving a >$10k load is the issue. > >Also, with OEM supplies you do not have access to the input side. Else >the warranty is toast. Unless qties are huge and you can negotiate that >"minor" inconvenience away. Not all issues are electronic ;-) > >Trying to short out a bank of serious electrolytics is IMHO not a smart >thing to do. First, they guarantee max ESR but usually never a min ESR. >Secondly, the trace resistance max can be calculated but not the min >because PCB fabs usually do not guarantee a max copper weight. You've >spec'd 2oz but what if it's a lot more? As an aside you'd also run the >risk of exploding an electrolytic and if something bad results from >that, oh boy, plaintiff's counsel would be all over you because I am >sure they can source the same state-of-the-art I have (schematics of >major supply mfgs) and use that as exhibits. > >Hint: There are emitter load share resistors and current sense resistors >which further limit the damage. Puff of smoke versus major kablouie ;-) > >>> >>>> Short series loop of input electrolytic and fuse and SCR... phzzzzt >>>> ;-) >>>> >>> And that's the problem ... the SCR needs to be huge. >> >> Again nonsense! Sounds like you've never actually designed a >> crowbar-protected supply ?? >> > >Au contraire. And they all tested out fine. I would never, ever, >consider placing a crowbar directly across a bank of serious input side >electrolytics. That is IMHO a recipe for disaster. > > >> I did my first 33 years ago, before simulation capability, worked >> perfectly ;-) >> >>> There is a really >>> fat one in the supply (>>TO220) but to get there would require major >>> surgery. >> >> Doesn't take a big "fat one". > > >It does. There is a good reason why they selected a much fatter one on >this 18A supply. If I could get at it I'd be home. And if there ain't >any COTS modules I'll have to. > >> >> The clue to your learning process is I^2*t >> > >I know about that. But business reality in systems architecture is often >like this: There is no fuse before the regulator, For crowbar use, there should be... makes for trivial protection of the, as you say, $BIG load ;-) >only one before the >transformer. Meaning the caps are free to dump their charge and good >ones do so with gusto. You can place a fuse between the crowbar and the >output of the supply but you must assume that it'll be an automotive >rated slow fuse and that the pass devices have fully fused shut. IOW >worst case, only load share and current sense resistors left in the game >until the fuse decides to finally blow. > >I have here various schematics of OEM linears from reputable US >manufacturers (unfortunately not at liberty to post on the web). All >have the SCR connected to the output rails, just like I always did. Now >all those manufacturers can't be wrong, can they? They're probably all Democrats ;-) ...Jim Thompson -- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine Sometimes I even put it in the food
From: Joerg on 18 Dec 2008 15:48 Jim Thompson wrote: > On Thu, 18 Dec 2008 11:16:38 -0800, Joerg > <notthisjoergsch(a)removethispacbell.net> wrote: > >> Jim Thompson wrote: >>> On Thu, 18 Dec 2008 09:53:45 -0800, Joerg >>> <notthisjoergsch(a)removethispacbell.net> wrote: >>> >>>> Jim Thompson wrote: >>>>> On Wed, 17 Dec 2008 15:44:48 -0800, Joerg >>>>> <notthisjoergsch(a)removethispacbell.net> wrote: >>>>> >>>>>> Hello All, >>>>>> >>>>>> After modding a 5V linear to 3.3V I feel a bit iffy about the crowbar, >>>>>> considering the Dollar amount that is going to be connected to it. The >>>>>> original crowbar is partially under (!) a big electrolytic, a bear to >>>>>> modify. Plus it probably isn't a good one to begin with, couldn't see a >>>>>> TL431 in there. >>>>>> >>>>>> So, is there a big old fat crowbar module anywhere that can wrestle an >>>>>> 18A linear supply? The usual OVP modules from Condor and others only >>>>>> have a smallish TO220 SCR in there which would probably go kablouie >>>>>> before it can protect anything. >>>>> Second question. You _are_ aware that the crowbar goes on the _input_ >>>>> side of the linear regulator, aren't you ?:-) >>>>> >>>> Ideally yes. Reality, no. Most crowbars are across the output terminals. >>>> They do their job because typically the supply is toast anyway if the >>>> crowbar has to kick in. IOW one of the pass transistors has fused shut >>>> and then it doesn't matter where the crowbar sits. >>> Nonsense! Regulator loop fails, pass device can't pass enough current >>> to blow fuse. >>> >> Not nonsense at all. Here's why: The resistance of the pass device, for >> whatever reason it is there, is desired. It makes sure the SCR doesn't >> turn into a plume of smoke. > > That statement proves you don't understand I^2*t > Again, show me a commercial grade linear supply from Condor, PowerOne or whatever that has an extra fuse between capacitor bank and pass device. With a good bank of caps well in excess of a total of 10,000uF a TO220 device doesn't stand much of a chance. You wouldn't even here phssst, just *KABLAM*. I will not do that. I assume we'll never reach agreement on this one. Anyhow, as far as I am concerned I'll continue to side with the design engineers at the major power supply manufacturers. >> The pass devices will likely blow but who >> cares? Saving the $0.50/each pass devices is not the issue, saving a >> $10k load is the issue. >> >> Also, with OEM supplies you do not have access to the input side. Else >> the warranty is toast. Unless qties are huge and you can negotiate that >> "minor" inconvenience away. Not all issues are electronic ;-) >> >> Trying to short out a bank of serious electrolytics is IMHO not a smart >> thing to do. First, they guarantee max ESR but usually never a min ESR. >> Secondly, the trace resistance max can be calculated but not the min >> because PCB fabs usually do not guarantee a max copper weight. You've >> spec'd 2oz but what if it's a lot more? As an aside you'd also run the >> risk of exploding an electrolytic and if something bad results from >> that, oh boy, plaintiff's counsel would be all over you because I am >> sure they can source the same state-of-the-art I have (schematics of >> major supply mfgs) and use that as exhibits. >> >> Hint: There are emitter load share resistors and current sense resistors >> which further limit the damage. Puff of smoke versus major kablouie ;-) >> >>>>> Short series loop of input electrolytic and fuse and SCR... phzzzzt >>>>> ;-) >>>>> >>>> And that's the problem ... the SCR needs to be huge. >>> Again nonsense! Sounds like you've never actually designed a >>> crowbar-protected supply ?? >>> >> Au contraire. And they all tested out fine. I would never, ever, >> consider placing a crowbar directly across a bank of serious input side >> electrolytics. That is IMHO a recipe for disaster. >> >> >>> I did my first 33 years ago, before simulation capability, worked >>> perfectly ;-) >>> >>>> There is a really >>>> fat one in the supply (>>TO220) but to get there would require major >>>> surgery. >>> Doesn't take a big "fat one". >> >> It does. There is a good reason why they selected a much fatter one on >> this 18A supply. If I could get at it I'd be home. And if there ain't >> any COTS modules I'll have to. >> >>> The clue to your learning process is I^2*t >>> >> I know about that. But business reality in systems architecture is often >> like this: There is no fuse before the regulator, > > For crowbar use, there should be... makes for trivial protection of > the, as you say, $BIG load ;-) > Hanging the crowbar onto the output makes it even more trivial. Which is why that's how it's done :-) >> only one before the >> transformer. Meaning the caps are free to dump their charge and good >> ones do so with gusto. You can place a fuse between the crowbar and the >> output of the supply but you must assume that it'll be an automotive >> rated slow fuse and that the pass devices have fully fused shut. IOW >> worst case, only load share and current sense resistors left in the game >> until the fuse decides to finally blow. >> >> I have here various schematics of OEM linears from reputable US >> manufacturers (unfortunately not at liberty to post on the web). All >> have the SCR connected to the output rails, just like I always did. Now >> all those manufacturers can't be wrong, can they? > > They're probably all Democrats ;-) > I hope none of them read that ;-) -- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
From: Jim Thompson on 18 Dec 2008 16:28
On Thu, 18 Dec 2008 12:48:54 -0800, Joerg <notthisjoergsch(a)removethispacbell.net> wrote: >Jim Thompson wrote: >> On Thu, 18 Dec 2008 11:16:38 -0800, Joerg >> <notthisjoergsch(a)removethispacbell.net> wrote: >> >>> Jim Thompson wrote: >>>> On Thu, 18 Dec 2008 09:53:45 -0800, Joerg >>>> <notthisjoergsch(a)removethispacbell.net> wrote: >>>> >>>>> Jim Thompson wrote: >>>>>> On Wed, 17 Dec 2008 15:44:48 -0800, Joerg >>>>>> <notthisjoergsch(a)removethispacbell.net> wrote: >>>>>> >>>>>>> Hello All, >>>>>>> >>>>>>> After modding a 5V linear to 3.3V I feel a bit iffy about the crowbar, >>>>>>> considering the Dollar amount that is going to be connected to it. The >>>>>>> original crowbar is partially under (!) a big electrolytic, a bear to >>>>>>> modify. Plus it probably isn't a good one to begin with, couldn't see a >>>>>>> TL431 in there. >>>>>>> >>>>>>> So, is there a big old fat crowbar module anywhere that can wrestle an >>>>>>> 18A linear supply? The usual OVP modules from Condor and others only >>>>>>> have a smallish TO220 SCR in there which would probably go kablouie >>>>>>> before it can protect anything. >>>>>> Second question. You _are_ aware that the crowbar goes on the _input_ >>>>>> side of the linear regulator, aren't you ?:-) >>>>>> >>>>> Ideally yes. Reality, no. Most crowbars are across the output terminals. >>>>> They do their job because typically the supply is toast anyway if the >>>>> crowbar has to kick in. IOW one of the pass transistors has fused shut >>>>> and then it doesn't matter where the crowbar sits. >>>> Nonsense! Regulator loop fails, pass device can't pass enough current >>>> to blow fuse. >>>> >>> Not nonsense at all. Here's why: The resistance of the pass device, for >>> whatever reason it is there, is desired. It makes sure the SCR doesn't >>> turn into a plume of smoke. >> >> That statement proves you don't understand I^2*t >> > >Again, show me a commercial grade linear supply from Condor, PowerOne or >whatever that has an extra fuse between capacitor bank and pass device. >With a good bank of caps well in excess of a total of 10,000uF a TO220 >device doesn't stand much of a chance. You wouldn't even here phssst, >just *KABLAM*. I will not do that. > >I assume we'll never reach agreement on this one. Anyhow, as far as I am >concerned I'll continue to side with the design engineers at the major >power supply manufacturers. > > >>> The pass devices will likely blow but who >>> cares? Saving the $0.50/each pass devices is not the issue, saving a >>> $10k load is the issue. >>> >>> Also, with OEM supplies you do not have access to the input side. Else >>> the warranty is toast. Unless qties are huge and you can negotiate that >>> "minor" inconvenience away. Not all issues are electronic ;-) >>> >>> Trying to short out a bank of serious electrolytics is IMHO not a smart >>> thing to do. First, they guarantee max ESR but usually never a min ESR. >>> Secondly, the trace resistance max can be calculated but not the min >>> because PCB fabs usually do not guarantee a max copper weight. You've >>> spec'd 2oz but what if it's a lot more? As an aside you'd also run the >>> risk of exploding an electrolytic and if something bad results from >>> that, oh boy, plaintiff's counsel would be all over you because I am >>> sure they can source the same state-of-the-art I have (schematics of >>> major supply mfgs) and use that as exhibits. >>> >>> Hint: There are emitter load share resistors and current sense resistors >>> which further limit the damage. Puff of smoke versus major kablouie ;-) >>> >>>>>> Short series loop of input electrolytic and fuse and SCR... phzzzzt >>>>>> ;-) >>>>>> >>>>> And that's the problem ... the SCR needs to be huge. >>>> Again nonsense! Sounds like you've never actually designed a >>>> crowbar-protected supply ?? >>>> >>> Au contraire. And they all tested out fine. I would never, ever, >>> consider placing a crowbar directly across a bank of serious input side >>> electrolytics. That is IMHO a recipe for disaster. >>> >>> >>>> I did my first 33 years ago, before simulation capability, worked >>>> perfectly ;-) >>>> >>>>> There is a really >>>>> fat one in the supply (>>TO220) but to get there would require major >>>>> surgery. >>>> Doesn't take a big "fat one". >>> >>> It does. There is a good reason why they selected a much fatter one on >>> this 18A supply. If I could get at it I'd be home. And if there ain't >>> any COTS modules I'll have to. >>> >>>> The clue to your learning process is I^2*t >>>> >>> I know about that. But business reality in systems architecture is often >>> like this: There is no fuse before the regulator, >> >> For crowbar use, there should be... makes for trivial protection of >> the, as you say, $BIG load ;-) >> > >Hanging the crowbar onto the output makes it even more trivial. Which is >why that's how it's done :-) > > >>> only one before the >>> transformer. Meaning the caps are free to dump their charge and good >>> ones do so with gusto. You can place a fuse between the crowbar and the >>> output of the supply but you must assume that it'll be an automotive >>> rated slow fuse "SLOW" ??? Dum, dee, dum, dum ;-) >>> and that the pass devices have fully fused shut. IOW >>> worst case, only load share and current sense resistors left in the game >>> until the fuse decides to finally blow. >>> >>> I have here various schematics of OEM linears from reputable US >>> manufacturers (unfortunately not at liberty to post on the web). All >>> have the SCR connected to the output rails, just like I always did. Now >>> all those manufacturers can't be wrong, can they? >> >> They're probably all Democrats ;-) >> > >I hope none of them read that ;-) I guess there's no convincing you that the crowbar goes (properly) before the linear, all the energy to blow the fuse (between capacitor bank and linear) comes from the capacitor bank, with properly sized fuse (fuseI^2*t << scrI^*2t), SCR doesn't even flinch. I personally consider crow-barring output, with no fuses except primary, to be down-right stupid. But it's not my problem... My finished designs ALWAYS work ;-) ...Jim Thompson -- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine Sometimes I even put it in the food |