On Mon, 11 Sep 2017 13:27:26 -0700, John Larkin wrote:> Well, there don't seem to be SOAR specs on zeners. I suppose I'll have > to blow some up.Hang a diode on it as a pass power transistor setup or hang a power transister across it and havd a power shunt zener.
On Wed, 13 Sep 2017 20:43:52 -0400, Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:>On 09/13/2017 07:52 PM, John Larkin wrote: >> On Wed, 13 Sep 2017 16:29:34 -0400, Phil Hobbs >> <pcdhSpamMeSenseless@electrooptical.net> wrote: >> >>> On 09/12/2017 01:58 PM, John Larkin wrote: >>>> On Tue, 12 Sep 2017 12:10:40 -0500, "Tim Williams" >>>> <tmoranwms@gmail.com> wrote: >>>> >>>>> "Phil Hobbs" <pcdhSpamMeSenseless@electrooptical.net> wrote in message >>>>> news:QfydnRoVYIpmXirEnZ2dnUU7-YfNnZ2d@supernews.com... >>>>>> That's plausible, since the TC of avalanche current is small and negative, >>>>>> so hot spots would avalanche slightly less. However, depending on the >>>>>> construction of the diode, there may be some fairly significant lateral >>>>>> resistance in the epi. >>>>> >>>>> That, and sensitivity to defects. Wasn't that the problem with the early >>>>> diodes (before 1N4007 and the like arrived), they worked majestically well >>>>> within ratings, but could be toasted at the drop of a hat from a little >>>>> overvoltage? >>>>> >>>>> The implication being, I guess, faults in the junction, or even more likely, >>>>> along its edge, causing uneven avalanche (or weirder things) and rapid >>>>> failure under adverse conditions. >>>>> >>>>> Speaking of, when were guard rings developed? That, and careful control of >>>>> surface states (purity, cleaning, and passivation), must've been critical >>>>> steps towards reliable parts. >>>> >>>> Early silicon diodes, and some still I think, were just diced from a >>>> big diffused wafer. Envision lots of edge damage. >>>> >>>> >>>> >>>>> >>>>> >>>>>> Something like a MELF package would reduce this, because the metal makes >>>>>> contact over the whole surface of the die. Maybe the metal contact covers >>>>>> the whole die in the SMT parts as well, I don't know. >>>>> >>>>> Hmm, interesting thought. Also good way to get heat out -- one terminal is >>>>> even in contact with the junction side (for epitaxy or one side diffusion). >>>>> The lead can share in the first tens of microseconds of heating. >>>>> >>>>> For pulses on the order of what JL's talking, it's all about die size, and a >>>>> little about what's immediately touching it. >>>>> >>>>> >>>>>> I haven't measured it myself, but I doubt that zener recovery is slow at >>>>>> all. In reverse bias there's no stored charge in the junction, and so >>>>>> nothing to shield out the applied field. >>>>> >>>>> Zener breakdown, per se, is a tunneling phenomenon, which I would think is >>>>> pretty darned fast, both on and off. Avalanche should have a tail, but it >>>>> might be so fast (i.e., charges sweep out "instantly" because of the high >>>>> field and fully depleted junction) that you can't tell. >>>>> >>>>> BJTs under avalanche can take tens of microseconds to recover, but part of >>>>> that is due to the three-layer design*. I'm not sure how much is comparable >>>>> between BJTs and diodes. >>>>> >>>>> (*Hmm, it would entirely come down to the shielding effect of the base >>>>> layer, no? And likewise, one should expect quicker recovery when smaller >>>>> R_BE, or a B-E discharge circuit, is used. I should test that.) >>>>> >>>>> Regular power diodes are susceptible to dynamic breakdown (applying a high >>>>> reverse voltage during / just after reverse recovery), and I would expect >>>>> zeners to be as well. At least there are very few situations, where you'd >>>>> need to worry about a zener diode's behavior as it goes suddenly from >>>>> forward to reverse bias. And with the higher doping (of most "zeners", >>>>> including true zeners, obviously), it would be that much harder to trigger. >>>> >>>> There is a second Grehkov effect: apply a high voltage across a diode, >>>> so fast that it forgets to conduct. When it finally wakes up, it >>>> breaks down in picoseconds. >>>> >>>>> >>>>> Oh hm, speaking of doping, do they do zener diode families by doping >>>>> density, diffusion time, or both to cover the range? >>>>> >>>>> Oh and and, power diodes are often PIN diodes, which gives a lot of room for >>>>> charge drift shenannigans. >>>> >>>> Right, that's the first Grekhov effect, the drift step-recovery diode, >>>> which typically uses PIN-structure high-voltage rectifier diodes. >>>> Combine that with the second one, the sneaky over-voltage thing, and >>>> you can get kilovolt edges at giant currents in picoseconds. >>> >>> And the FCC will have no trouble finding you. ;) >>> >>> Cheers >>> >>> Phil Hobbs >>>> >>>> >> >> https://www.dropbox.com/s/bwaulog6mzx8zkn/T222_Copper.jpg?raw=1 >> >> https://www.dropbox.com/s/q82toc257fv43z8/DSRD_neg-2KV.JPG?raw=1 >> >> Water cooled, -2 kV pulse, 1.4 ns wide, 100 KHz. That's the DSRD >> version without the breakdown booster. That was for a tomographic atom >> probe project. > >One of these days I'd like to see that running, if you still have it. > >Cheers > >Phil HobbsI don't think I do. Haven't seen it in years... might be in storage somewhere. I wound up using the C-B junction of a 1500V TV horizontal-output transistor as the DSRD. The doping profile was accidentally right. That transistor is long gone. -- John Larkin Highland Technology, Inc lunatic fringe electronics
On 09/13/2017 07:52 PM, John Larkin wrote:> On Wed, 13 Sep 2017 16:29:34 -0400, Phil Hobbs > <pcdhSpamMeSenseless@electrooptical.net> wrote: > >> On 09/12/2017 01:58 PM, John Larkin wrote: >>> On Tue, 12 Sep 2017 12:10:40 -0500, "Tim Williams" >>> <tmoranwms@gmail.com> wrote: >>> >>>> "Phil Hobbs" <pcdhSpamMeSenseless@electrooptical.net> wrote in message >>>> news:QfydnRoVYIpmXirEnZ2dnUU7-YfNnZ2d@supernews.com... >>>>> That's plausible, since the TC of avalanche current is small and negative, >>>>> so hot spots would avalanche slightly less. However, depending on the >>>>> construction of the diode, there may be some fairly significant lateral >>>>> resistance in the epi. >>>> >>>> That, and sensitivity to defects. Wasn't that the problem with the early >>>> diodes (before 1N4007 and the like arrived), they worked majestically well >>>> within ratings, but could be toasted at the drop of a hat from a little >>>> overvoltage? >>>> >>>> The implication being, I guess, faults in the junction, or even more likely, >>>> along its edge, causing uneven avalanche (or weirder things) and rapid >>>> failure under adverse conditions. >>>> >>>> Speaking of, when were guard rings developed? That, and careful control of >>>> surface states (purity, cleaning, and passivation), must've been critical >>>> steps towards reliable parts. >>> >>> Early silicon diodes, and some still I think, were just diced from a >>> big diffused wafer. Envision lots of edge damage. >>> >>> >>> >>>> >>>> >>>>> Something like a MELF package would reduce this, because the metal makes >>>>> contact over the whole surface of the die. Maybe the metal contact covers >>>>> the whole die in the SMT parts as well, I don't know. >>>> >>>> Hmm, interesting thought. Also good way to get heat out -- one terminal is >>>> even in contact with the junction side (for epitaxy or one side diffusion). >>>> The lead can share in the first tens of microseconds of heating. >>>> >>>> For pulses on the order of what JL's talking, it's all about die size, and a >>>> little about what's immediately touching it. >>>> >>>> >>>>> I haven't measured it myself, but I doubt that zener recovery is slow at >>>>> all. In reverse bias there's no stored charge in the junction, and so >>>>> nothing to shield out the applied field. >>>> >>>> Zener breakdown, per se, is a tunneling phenomenon, which I would think is >>>> pretty darned fast, both on and off. Avalanche should have a tail, but it >>>> might be so fast (i.e., charges sweep out "instantly" because of the high >>>> field and fully depleted junction) that you can't tell. >>>> >>>> BJTs under avalanche can take tens of microseconds to recover, but part of >>>> that is due to the three-layer design*. I'm not sure how much is comparable >>>> between BJTs and diodes. >>>> >>>> (*Hmm, it would entirely come down to the shielding effect of the base >>>> layer, no? And likewise, one should expect quicker recovery when smaller >>>> R_BE, or a B-E discharge circuit, is used. I should test that.) >>>> >>>> Regular power diodes are susceptible to dynamic breakdown (applying a high >>>> reverse voltage during / just after reverse recovery), and I would expect >>>> zeners to be as well. At least there are very few situations, where you'd >>>> need to worry about a zener diode's behavior as it goes suddenly from >>>> forward to reverse bias. And with the higher doping (of most "zeners", >>>> including true zeners, obviously), it would be that much harder to trigger. >>> >>> There is a second Grehkov effect: apply a high voltage across a diode, >>> so fast that it forgets to conduct. When it finally wakes up, it >>> breaks down in picoseconds. >>> >>>> >>>> Oh hm, speaking of doping, do they do zener diode families by doping >>>> density, diffusion time, or both to cover the range? >>>> >>>> Oh and and, power diodes are often PIN diodes, which gives a lot of room for >>>> charge drift shenannigans. >>> >>> Right, that's the first Grekhov effect, the drift step-recovery diode, >>> which typically uses PIN-structure high-voltage rectifier diodes. >>> Combine that with the second one, the sneaky over-voltage thing, and >>> you can get kilovolt edges at giant currents in picoseconds. >> >> And the FCC will have no trouble finding you. ;) >> >> Cheers >> >> Phil Hobbs >>> >>> > > https://www.dropbox.com/s/bwaulog6mzx8zkn/T222_Copper.jpg?raw=1 > > https://www.dropbox.com/s/q82toc257fv43z8/DSRD_neg-2KV.JPG?raw=1 > > Water cooled, -2 kV pulse, 1.4 ns wide, 100 KHz. That's the DSRD > version without the breakdown booster. That was for a tomographic atom > probe project.One of these days I'd like to see that running, if you still have it. Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
On Wed, 13 Sep 2017 16:29:34 -0400, Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:>On 09/12/2017 01:58 PM, John Larkin wrote: >> On Tue, 12 Sep 2017 12:10:40 -0500, "Tim Williams" >> <tmoranwms@gmail.com> wrote: >> >>> "Phil Hobbs" <pcdhSpamMeSenseless@electrooptical.net> wrote in message >>> news:QfydnRoVYIpmXirEnZ2dnUU7-YfNnZ2d@supernews.com... >>>> That's plausible, since the TC of avalanche current is small and negative, >>>> so hot spots would avalanche slightly less. However, depending on the >>>> construction of the diode, there may be some fairly significant lateral >>>> resistance in the epi. >>> >>> That, and sensitivity to defects. Wasn't that the problem with the early >>> diodes (before 1N4007 and the like arrived), they worked majestically well >>> within ratings, but could be toasted at the drop of a hat from a little >>> overvoltage? >>> >>> The implication being, I guess, faults in the junction, or even more likely, >>> along its edge, causing uneven avalanche (or weirder things) and rapid >>> failure under adverse conditions. >>> >>> Speaking of, when were guard rings developed? That, and careful control of >>> surface states (purity, cleaning, and passivation), must've been critical >>> steps towards reliable parts. >> >> Early silicon diodes, and some still I think, were just diced from a >> big diffused wafer. Envision lots of edge damage. >> >> >> >>> >>> >>>> Something like a MELF package would reduce this, because the metal makes >>>> contact over the whole surface of the die. Maybe the metal contact covers >>>> the whole die in the SMT parts as well, I don't know. >>> >>> Hmm, interesting thought. Also good way to get heat out -- one terminal is >>> even in contact with the junction side (for epitaxy or one side diffusion). >>> The lead can share in the first tens of microseconds of heating. >>> >>> For pulses on the order of what JL's talking, it's all about die size, and a >>> little about what's immediately touching it. >>> >>> >>>> I haven't measured it myself, but I doubt that zener recovery is slow at >>>> all. In reverse bias there's no stored charge in the junction, and so >>>> nothing to shield out the applied field. >>> >>> Zener breakdown, per se, is a tunneling phenomenon, which I would think is >>> pretty darned fast, both on and off. Avalanche should have a tail, but it >>> might be so fast (i.e., charges sweep out "instantly" because of the high >>> field and fully depleted junction) that you can't tell. >>> >>> BJTs under avalanche can take tens of microseconds to recover, but part of >>> that is due to the three-layer design*. I'm not sure how much is comparable >>> between BJTs and diodes. >>> >>> (*Hmm, it would entirely come down to the shielding effect of the base >>> layer, no? And likewise, one should expect quicker recovery when smaller >>> R_BE, or a B-E discharge circuit, is used. I should test that.) >>> >>> Regular power diodes are susceptible to dynamic breakdown (applying a high >>> reverse voltage during / just after reverse recovery), and I would expect >>> zeners to be as well. At least there are very few situations, where you'd >>> need to worry about a zener diode's behavior as it goes suddenly from >>> forward to reverse bias. And with the higher doping (of most "zeners", >>> including true zeners, obviously), it would be that much harder to trigger. >> >> There is a second Grehkov effect: apply a high voltage across a diode, >> so fast that it forgets to conduct. When it finally wakes up, it >> breaks down in picoseconds. >> >>> >>> Oh hm, speaking of doping, do they do zener diode families by doping >>> density, diffusion time, or both to cover the range? >>> >>> Oh and and, power diodes are often PIN diodes, which gives a lot of room for >>> charge drift shenannigans. >> >> Right, that's the first Grekhov effect, the drift step-recovery diode, >> which typically uses PIN-structure high-voltage rectifier diodes. >> Combine that with the second one, the sneaky over-voltage thing, and >> you can get kilovolt edges at giant currents in picoseconds. > >And the FCC will have no trouble finding you. ;) > >Cheers > >Phil Hobbs >> >>https://www.dropbox.com/s/bwaulog6mzx8zkn/T222_Copper.jpg?raw=1 https://www.dropbox.com/s/q82toc257fv43z8/DSRD_neg-2KV.JPG?raw=1 Water cooled, -2 kV pulse, 1.4 ns wide, 100 KHz. That's the DSRD version without the breakdown booster. That was for a tomographic atom probe project. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
On 09/12/2017 01:58 PM, John Larkin wrote:> On Tue, 12 Sep 2017 12:10:40 -0500, "Tim Williams" > <tmoranwms@gmail.com> wrote: > >> "Phil Hobbs" <pcdhSpamMeSenseless@electrooptical.net> wrote in message >> news:QfydnRoVYIpmXirEnZ2dnUU7-YfNnZ2d@supernews.com... >>> That's plausible, since the TC of avalanche current is small and negative, >>> so hot spots would avalanche slightly less. However, depending on the >>> construction of the diode, there may be some fairly significant lateral >>> resistance in the epi. >> >> That, and sensitivity to defects. Wasn't that the problem with the early >> diodes (before 1N4007 and the like arrived), they worked majestically well >> within ratings, but could be toasted at the drop of a hat from a little >> overvoltage? >> >> The implication being, I guess, faults in the junction, or even more likely, >> along its edge, causing uneven avalanche (or weirder things) and rapid >> failure under adverse conditions. >> >> Speaking of, when were guard rings developed? That, and careful control of >> surface states (purity, cleaning, and passivation), must've been critical >> steps towards reliable parts. > > Early silicon diodes, and some still I think, were just diced from a > big diffused wafer. Envision lots of edge damage. > > > >> >> >>> Something like a MELF package would reduce this, because the metal makes >>> contact over the whole surface of the die. Maybe the metal contact covers >>> the whole die in the SMT parts as well, I don't know. >> >> Hmm, interesting thought. Also good way to get heat out -- one terminal is >> even in contact with the junction side (for epitaxy or one side diffusion). >> The lead can share in the first tens of microseconds of heating. >> >> For pulses on the order of what JL's talking, it's all about die size, and a >> little about what's immediately touching it. >> >> >>> I haven't measured it myself, but I doubt that zener recovery is slow at >>> all. In reverse bias there's no stored charge in the junction, and so >>> nothing to shield out the applied field. >> >> Zener breakdown, per se, is a tunneling phenomenon, which I would think is >> pretty darned fast, both on and off. Avalanche should have a tail, but it >> might be so fast (i.e., charges sweep out "instantly" because of the high >> field and fully depleted junction) that you can't tell. >> >> BJTs under avalanche can take tens of microseconds to recover, but part of >> that is due to the three-layer design*. I'm not sure how much is comparable >> between BJTs and diodes. >> >> (*Hmm, it would entirely come down to the shielding effect of the base >> layer, no? And likewise, one should expect quicker recovery when smaller >> R_BE, or a B-E discharge circuit, is used. I should test that.) >> >> Regular power diodes are susceptible to dynamic breakdown (applying a high >> reverse voltage during / just after reverse recovery), and I would expect >> zeners to be as well. At least there are very few situations, where you'd >> need to worry about a zener diode's behavior as it goes suddenly from >> forward to reverse bias. And with the higher doping (of most "zeners", >> including true zeners, obviously), it would be that much harder to trigger. > > There is a second Grehkov effect: apply a high voltage across a diode, > so fast that it forgets to conduct. When it finally wakes up, it > breaks down in picoseconds. > >> >> Oh hm, speaking of doping, do they do zener diode families by doping >> density, diffusion time, or both to cover the range? >> >> Oh and and, power diodes are often PIN diodes, which gives a lot of room for >> charge drift shenannigans. > > Right, that's the first Grekhov effect, the drift step-recovery diode, > which typically uses PIN-structure high-voltage rectifier diodes. > Combine that with the second one, the sneaky over-voltage thing, and > you can get kilovolt edges at giant currents in picoseconds.And the FCC will have no trouble finding you. ;) Cheers Phil Hobbs> >-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
On Tuesday, September 12, 2017 at 10:20:11 AM UTC-4, John Larkin wrote:> On Tue, 12 Sep 2017 07:24:43 -0400, Phil Hobbs > <pcdhSpamMeSenseless@electrooptical.net> wrote: > > >On 09/11/2017 11:36 PM, Tim Williams wrote: > >> There's an old Motorola (now On Semi) appnote talking about zener > >> ratings for pulse, ESD and surge purposes. > >> > >> While few zeners carry pulse ratings, there doesn't seem to be any > >> reason to suspect any average part is exceptionally inferior to those > >> documented in the appnote. > >> > >> TVSs are just big fat zeners with pulse ratings rather than DC ratings. > >> > >> I don't know of any mechanism that would resemble 2nd breakdown, or give > >> any more than a straight linear SOA (i.e., straight downward-sloping > >> lines on the log-log plot, with different offsets for different > >> single-shot pulse widths). > >That's plausible, since the TC of avalanche current is small and > >negative, so hot spots would avalanche slightly less. However, > >depending on the construction of the diode, there may be some fairly > >significant lateral resistance in the epi. > > > >Something like a MELF package would reduce this, because the metal makes > >contact over the whole surface of the die. Maybe the metal contact > >covers the whole die in the SMT parts as well, I don't know. > >> > >> As a diode, the usual charge storage and recovery business applies, but > >> you wouldn't normally use a zener as a rectifier, or under fast recovery > >> or drift-recovery conditions, so that should be easy to avoid. > > > >I haven't measured it myself, but I doubt that zener recovery is slow at > >all. In reverse bias there's no stored charge in the junction, and so > >nothing to shield out the applied field. > > > >Cheers > > > >Phil Hobbs > > First few zener data sheets that I liiked at, there were just DC power > specs. Then I found some others, SOT23 sizes, that have power:time > curves, typical value around 30 watts for 1 millisecond. That's pretty > impressive. > > I only need enough to turn a non-avalanche-controlled fet into one > that clamps safely in the 100 volt ballpark, which it looks like the > smallest zeners can do without a problem. The zener only needs to > furnish 50 nC or so of gate charge.So you're sticking the zener from the drain to the gate? Do you have to worry about the zener leakage current? I guess a fairly stiff drive voltage. George H.> > Avalanche-rated mosfets can typically clamp tens of mJ. Looks like the > external zener version can gulp a lot more. > > > -- > > John Larkin Highland Technology, Inc > > lunatic fringe electronics
On Tue, 12 Sep 2017 07:35:00 -0700, Joerg <news@analogconsultants.com> wrote:>On 2017-09-11 15:36, John Larkin wrote: >> On Mon, 11 Sep 2017 14:50:32 -0700, Joerg <news@analogconsultants.com> >> wrote: >> >>> On 2017-09-11 14:38, John Larkin wrote: >>>> On Mon, 11 Sep 2017 13:40:36 -0700, Joerg <news@analogconsultants.com> >>>> wrote: >>>> >>>>> On 2017-09-11 13:27, John Larkin wrote: >>>>>> >>>>>> Well, there don't seem to be SOAR specs on zeners. I suppose I'll have >>>>>> to blow some up. >>>>>> >>>>> >>>>> Why should there be? A zener either conducts next to nothing or it >>>>> starts to conduct and then the voltage drop is always very close to its >>>>> zener voltage or a diode drop when in the other direction. They do give >>>>> power specs. >>>> >>>> The issue is how it survives transient power dissipation. You could >>>> say "never exceed the continuous power rating" but then you could say >>>> the same for mosfets, and not have SOAR curves. >>>> >>>> For a zener, it would amount to a power-vs-time or a current-vs-time >>>> curve. >>>> >>>> Transzorbs usually have a SOAR curve or some equivalent. Some SMB >>>> packaged surface-mount transzorbs can dissipate 600 watts for a while. >>>> >>> >>> All you get with zeners is a thermal response graph set like figure 4 here: >>> >>> http://www.onsemi.com/pub/Collateral/1N5333B-D.PDF >>> >>> I am not in any way an expert on this but super-high pulse dissipation >>> is AFAIK not the domain where regular zeners would shine. Whenever I >>> needed that I made active zeners where a FET took over the wrestling >>> work. TVS are ok as well but they often have tolerances that are too >>> high and too much capacitance. >> >> >> Some zeners specify transient power limits, some don't. Some SOT23s >> can dissipate 30 watts for a millisecond, impressive. >> > >That's sounds like those insane "rated power levels" for some FETs or >PMPO audio amp ratings.It's basically a 1-dimensional SOAR curve, allowable power vs time. I need tens of microjoules and it looks like even the smallest zeners are spec'd to absorb millijoules. Of course, we've detonated mosfets within their SOAR curves, but I should be OK with a 1000:1 energy margin. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
On Tue, 12 Sep 2017 12:10:40 -0500, "Tim Williams" <tmoranwms@gmail.com> wrote:>"Phil Hobbs" <pcdhSpamMeSenseless@electrooptical.net> wrote in message >news:QfydnRoVYIpmXirEnZ2dnUU7-YfNnZ2d@supernews.com... >> That's plausible, since the TC of avalanche current is small and negative, >> so hot spots would avalanche slightly less. However, depending on the >> construction of the diode, there may be some fairly significant lateral >> resistance in the epi. > >That, and sensitivity to defects. Wasn't that the problem with the early >diodes (before 1N4007 and the like arrived), they worked majestically well >within ratings, but could be toasted at the drop of a hat from a little >overvoltage? > >The implication being, I guess, faults in the junction, or even more likely, >along its edge, causing uneven avalanche (or weirder things) and rapid >failure under adverse conditions. > >Speaking of, when were guard rings developed? That, and careful control of >surface states (purity, cleaning, and passivation), must've been critical >steps towards reliable parts.Early silicon diodes, and some still I think, were just diced from a big diffused wafer. Envision lots of edge damage.> > >> Something like a MELF package would reduce this, because the metal makes >> contact over the whole surface of the die. Maybe the metal contact covers >> the whole die in the SMT parts as well, I don't know. > >Hmm, interesting thought. Also good way to get heat out -- one terminal is >even in contact with the junction side (for epitaxy or one side diffusion). >The lead can share in the first tens of microseconds of heating. > >For pulses on the order of what JL's talking, it's all about die size, and a >little about what's immediately touching it. > > >> I haven't measured it myself, but I doubt that zener recovery is slow at >> all. In reverse bias there's no stored charge in the junction, and so >> nothing to shield out the applied field. > >Zener breakdown, per se, is a tunneling phenomenon, which I would think is >pretty darned fast, both on and off. Avalanche should have a tail, but it >might be so fast (i.e., charges sweep out "instantly" because of the high >field and fully depleted junction) that you can't tell. > >BJTs under avalanche can take tens of microseconds to recover, but part of >that is due to the three-layer design*. I'm not sure how much is comparable >between BJTs and diodes. > >(*Hmm, it would entirely come down to the shielding effect of the base >layer, no? And likewise, one should expect quicker recovery when smaller >R_BE, or a B-E discharge circuit, is used. I should test that.) > >Regular power diodes are susceptible to dynamic breakdown (applying a high >reverse voltage during / just after reverse recovery), and I would expect >zeners to be as well. At least there are very few situations, where you'd >need to worry about a zener diode's behavior as it goes suddenly from >forward to reverse bias. And with the higher doping (of most "zeners", >including true zeners, obviously), it would be that much harder to trigger.There is a second Grehkov effect: apply a high voltage across a diode, so fast that it forgets to conduct. When it finally wakes up, it breaks down in picoseconds.> >Oh hm, speaking of doping, do they do zener diode families by doping >density, diffusion time, or both to cover the range? > >Oh and and, power diodes are often PIN diodes, which gives a lot of room for >charge drift shenannigans.Right, that's the first Grekhov effect, the drift step-recovery diode, which typically uses PIN-structure high-voltage rectifier diodes. Combine that with the second one, the sneaky over-voltage thing, and you can get kilovolt edges at giant currents in picoseconds. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
On Tue, 12 Sep 2017 07:20:01 -0700, John Larkin <jjlarkin@highlandtechnology.com> wrote:>On Tue, 12 Sep 2017 07:24:43 -0400, Phil Hobbs ><pcdhSpamMeSenseless@electrooptical.net> wrote: > >>On 09/11/2017 11:36 PM, Tim Williams wrote: >>> There's an old Motorola (now On Semi) appnote talking about zener >>> ratings for pulse, ESD and surge purposes. >>> >>> While few zeners carry pulse ratings, there doesn't seem to be any >>> reason to suspect any average part is exceptionally inferior to those >>> documented in the appnote. >>> >>> TVSs are just big fat zeners with pulse ratings rather than DC ratings. >>> >>> I don't know of any mechanism that would resemble 2nd breakdown, or give >>> any more than a straight linear SOA (i.e., straight downward-sloping >>> lines on the log-log plot, with different offsets for different >>> single-shot pulse widths). >>That's plausible, since the TC of avalanche current is small and >>negative, so hot spots would avalanche slightly less. However, >>depending on the construction of the diode, there may be some fairly >>significant lateral resistance in the epi. >> >>Something like a MELF package would reduce this, because the metal makes >>contact over the whole surface of the die. Maybe the metal contact >>covers the whole die in the SMT parts as well, I don't know. >>> >>> As a diode, the usual charge storage and recovery business applies, but >>> you wouldn't normally use a zener as a rectifier, or under fast recovery >>> or drift-recovery conditions, so that should be easy to avoid. >> >>I haven't measured it myself, but I doubt that zener recovery is slow at >>all. In reverse bias there's no stored charge in the junction, and so >>nothing to shield out the applied field. >> >>Cheers >> >>Phil Hobbs > >First few zener data sheets that I liiked at, there were just DC power >specs. Then I found some others, SOT23 sizes, that have power:time >curves, typical value around 30 watts for 1 millisecond. That's pretty >impressive. > >I only need enough to turn a non-avalanche-controlled fet into one >that clamps safely in the 100 volt ballpark, which it looks like the >smallest zeners can do without a problem. The zener only needs to >furnish 50 nC or so of gate charge. > >Avalanche-rated mosfets can typically clamp tens of mJ. Looks like the >external zener version can gulp a lot more.Rob here looked into the theory of avalanche-rated fets. They essentially add a smallish transistor in parallel with the main mosfet - not to the gate - and it gets the avalanche energy. So the avalanche energy is way below the SOAR curve energy. The external zener circuit uses the main fet to dump energy, so it can absorb a lot more energy, ballpark 10x the avalanche rating. Plus, the zener works for any fet, avalanche rated or not. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
"Phil Hobbs" <pcdhSpamMeSenseless@electrooptical.net> wrote in message news:QfydnRoVYIpmXirEnZ2dnUU7-YfNnZ2d@supernews.com...> That's plausible, since the TC of avalanche current is small and negative, > so hot spots would avalanche slightly less. However, depending on the > construction of the diode, there may be some fairly significant lateral > resistance in the epi.That, and sensitivity to defects. Wasn't that the problem with the early diodes (before 1N4007 and the like arrived), they worked majestically well within ratings, but could be toasted at the drop of a hat from a little overvoltage? The implication being, I guess, faults in the junction, or even more likely, along its edge, causing uneven avalanche (or weirder things) and rapid failure under adverse conditions. Speaking of, when were guard rings developed? That, and careful control of surface states (purity, cleaning, and passivation), must've been critical steps towards reliable parts.> Something like a MELF package would reduce this, because the metal makes > contact over the whole surface of the die. Maybe the metal contact covers > the whole die in the SMT parts as well, I don't know.Hmm, interesting thought. Also good way to get heat out -- one terminal is even in contact with the junction side (for epitaxy or one side diffusion). The lead can share in the first tens of microseconds of heating. For pulses on the order of what JL's talking, it's all about die size, and a little about what's immediately touching it.> I haven't measured it myself, but I doubt that zener recovery is slow at > all. In reverse bias there's no stored charge in the junction, and so > nothing to shield out the applied field.Zener breakdown, per se, is a tunneling phenomenon, which I would think is pretty darned fast, both on and off. Avalanche should have a tail, but it might be so fast (i.e., charges sweep out "instantly" because of the high field and fully depleted junction) that you can't tell. BJTs under avalanche can take tens of microseconds to recover, but part of that is due to the three-layer design*. I'm not sure how much is comparable between BJTs and diodes. (*Hmm, it would entirely come down to the shielding effect of the base layer, no? And likewise, one should expect quicker recovery when smaller R_BE, or a B-E discharge circuit, is used. I should test that.) Regular power diodes are susceptible to dynamic breakdown (applying a high reverse voltage during / just after reverse recovery), and I would expect zeners to be as well. At least there are very few situations, where you'd need to worry about a zener diode's behavior as it goes suddenly from forward to reverse bias. And with the higher doping (of most "zeners", including true zeners, obviously), it would be that much harder to trigger. Oh hm, speaking of doping, do they do zener diode families by doping density, diffusion time, or both to cover the range? Oh and and, power diodes are often PIN diodes, which gives a lot of room for charge drift shenannigans. It occurs to me, zeners are less abundant over 200V, which may be a market thing (who needs 'em / use MOVs?), but also may be a physics thing. A PIN avalanche diode might not be as desirable, and it seems to me, around 200V is where PIN strutures start being preferable for rectifiers. Tim -- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com