John Larkin>On Fri, 12 Oct 2018 14:01:14 -0700, John Larkin ><jjlarkin@highland_snip_technology.com> wrote: > >> >> >>I'm still iterating on my high-voltage pulse generator. Rev C looks >>very good. I've been pushing it up voltage and frequency (like, 1200 >>volts and 4 MHz) and seeing what parts fry, and working on them. >> >>I have a dpak schottky diode in a sort of snubber place. According to >>Spice, it shouldn't get too hot, and it's soldered to all sorts of >>copper pours and thermal vias and stuff. It's an SBRD10200 and it's >>frying. I happened to have some smaller, lower current diodes around, >>UPS5100E3, which is a sawed-off sub-dpak thing. I figured I might >>squeeze in two of those in parallel... but just one runs cool. >> >>https://www.dropbox.com/s/6zv36c4d6v7dsbn/Two_Diodes.JPG?raw=1 >> >>Weird. I'm guessing that the big one has a p-n guard ring that makes >>it have reverse recovery current right after a big forward current >>spike. Data sheets never talk about that. >> >>I heat sinked things too well, so it's a bear to desolder/resolder. >> >>https://www.dropbox.com/s/bg0399wzzi4k3ze/T850C_D1_Kluge.jpg?raw=1 > > >Replacing three schottky diodes with the cute little Microsemi parts >helped my pulse generator efficiency a lot. They also solder into an >SMB diode footprint, with some care to not short out. > >My circuit still isn't as efficient as the Spice sim, so I've been >prowling around with our FLIR imager to see where the heat is. >Surprisingly (to me) at 1KV and 4 MHz pulse rate, the Phoenix output >terminal block is hot. > >I ripped out the metal bits from the unused center terminal, and it >really cooled off. > >https://www.dropbox.com/s/ogiwi79ovipishe/T850_Phoenix.JPG?raw=1 > >Power consumption dropped by 3.4 watts. > >I suppose the ham radio RF power guys already knew about stuff like >this; I'm still learning.Cool, great, yes those are not really RF blocks... You should in, a free moment, read one of those old RSGB handbooks. All about construction of RF circuits up to some GHz. I sure learned a lot from that, as a kid. https://rsgbshop.org/acatalog/Online_Catalogue_Radio_Books___CDs_29.html No longer have the old RSGB handbook, mine from the tube era, but still can visualize pages from it... Spice, there was no spice, you measure things.. if you can...
guard ring?
Started by ●October 12, 2018
Reply by ●October 15, 20182018-10-15
Reply by ●October 15, 20182018-10-15
PS, those terminal blocks are not really that good, I get bad contact at DC with those too, finally soldered the wires: http://panteltje.com/pub/bad_terminal_blocks_IMG_6619.JPG Maybe one should not use tinned wire, but I do have high resistance effect. So it needs not be a RF problem, could just be resistive heating. Use the good old amphenol PL259 connectors for high power RF:-) https://www.newark.com/amphenol-rf/pl-259/connector-uhf-plug-cable/dp/37K8160
Reply by ●October 15, 20182018-10-15
Jasen Betts <jasen@xnet.co.nz> wrote in news:pq183r$40d$2@gonzo.alcatraz:> > No. And how the fuck would a POTS connection cause dropbox to fail > anyway? sometimes drop box is drop ball. > >Fail? No. But folks like you get impatient and start pointing fingers. So, POTS didn't do it... an impatient blametard did.
Reply by ●October 15, 20182018-10-15
<698839253X6D445TD@nospam.org> wrote in news:pq1h3u$1133$1@gioia.aioe.org:> PS, > those terminal blocks are not really that good, > I get bad contact at DC with those too, finally soldered the wires: > http://panteltje.com/pub/bad_terminal_blocks_IMG_6619.JPGThose are ideally meant for a solid wire. Their original design debut was. For stranded, one needs to crimp on a pin to the wire end or fudge that by tinning a segment of it, but then one must read up on the possibilities of a fail mode due to "solder creep". Generally not for small signall, but for power, solder creep can be an issue.> > Maybe one should not use tinned wire, but I do have high resistanceAlso high strand counf SPC wire instead of that cheap, low strand cound, bad copper, pvc coated crap wire there, you would get better results as well. SPC teflon sheathed wire spools are pretty cheap.> effect. So it needs not be a RF problem, could just be resistive > heating. Use the good old amphenol PL259 connectors for high power > RF:-) > https://www.newark.com/amphenol-rf/pl-259/connector-uhf-plug-cable/dp/> 37K8160 > > >
Reply by ●October 15, 20182018-10-15
On Mon, 15 Oct 2018 11:21:24 +0000 (UTC), DecadentLinuxUserNumeroUno@decadence.org wrote:><698839253X6D445TD@nospam.org> wrote in >news:pq1h3u$1133$1@gioia.aioe.org: > >> PS, >> those terminal blocks are not really that good, >> I get bad contact at DC with those too, finally soldered the wires: >> http://panteltje.com/pub/bad_terminal_blocks_IMG_6619.JPGThe Phoenix ones are fine. Some cheap ones are, well, cheap. Thermoplastics will flow and loosen up, which can be thermal runaway.> > Those are ideally meant for a solid wire. Their original design debut >was. For stranded, one needs to crimp on a pin to the wire end or fudge >that by tinning a segment of it, but then one must read up on the >possibilities of a fail mode due to "solder creep". Generally not for >small signall, but for power, solder creep can be an issue.My customer's gadget uses stranded wire. Those little Phoenix blocks have a very well machined (and hard to remove!) clamp mechanism that works well on stranded wire. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●October 16, 20182018-10-16
On Fri, 12 Oct 2018 14:01:14 -0700, John Larkin <jjlarkin@highland_snip_technology.com> wrote:> > >I'm still iterating on my high-voltage pulse generator. Rev C looks >very good. I've been pushing it up voltage and frequency (like, 1200 >volts and 4 MHz) and seeing what parts fry, and working on them. > >I have a dpak schottky diode in a sort of snubber place. According to >Spice, it shouldn't get too hot, and it's soldered to all sorts of >copper pours and thermal vias and stuff. It's an SBRD10200 and it's >frying. I happened to have some smaller, lower current diodes around, >UPS5100E3, which is a sawed-off sub-dpak thing. I figured I might >squeeze in two of those in parallel... but just one runs cool. > >https://www.dropbox.com/s/6zv36c4d6v7dsbn/Two_Diodes.JPG?raw=1 > >Weird. I'm guessing that the big one has a p-n guard ring that makes >it have reverse recovery current right after a big forward current >spike. Data sheets never talk about that. > >I heat sinked things too well, so it's a bear to desolder/resolder. > >https://www.dropbox.com/s/bg0399wzzi4k3ze/T850C_D1_Kluge.jpg?raw=1I may have discovered a weird thermal runaway mechanism. Possibly as temperature increases, more current shifts from the schottky junction into the guard ring, reverse recovery dissipation increases, and boom. I might try some silicon carbide. More Vf but no guard ring. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Reply by ●October 16, 20182018-10-16
On 10/16/18 1:31 PM, John Larkin wrote:> On Fri, 12 Oct 2018 14:01:14 -0700, John Larkin > <jjlarkin@highland_snip_technology.com> wrote: > >> >> >> I'm still iterating on my high-voltage pulse generator. Rev C looks >> very good. I've been pushing it up voltage and frequency (like, 1200 >> volts and 4 MHz) and seeing what parts fry, and working on them. >> >> I have a dpak schottky diode in a sort of snubber place. According to >> Spice, it shouldn't get too hot, and it's soldered to all sorts of >> copper pours and thermal vias and stuff. It's an SBRD10200 and it's >> frying. I happened to have some smaller, lower current diodes around, >> UPS5100E3, which is a sawed-off sub-dpak thing. I figured I might >> squeeze in two of those in parallel... but just one runs cool. >> >> https://www.dropbox.com/s/6zv36c4d6v7dsbn/Two_Diodes.JPG?raw=1 >> >> Weird. I'm guessing that the big one has a p-n guard ring that makes >> it have reverse recovery current right after a big forward current >> spike. Data sheets never talk about that. >> >> I heat sinked things too well, so it's a bear to desolder/resolder. >> >> https://www.dropbox.com/s/bg0399wzzi4k3ze/T850C_D1_Kluge.jpg?raw=1 > > I may have discovered a weird thermal runaway mechanism. Possibly as > temperature increases, more current shifts from the schottky junction > into the guard ring, reverse recovery dissipation increases, and boom.Wouldn't be surprised. Schottkys don't have the same tempco as PN diodes.> > I might try some silicon carbide. More Vf but no guard ring. > >Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC / Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics Briarcliff Manor NY 10510 http://electrooptical.net http://hobbs-eo.com
Reply by ●October 16, 20182018-10-16
"John Larkin" <jjlarkin@highland_snip_technology.com> wrote in message news:7r7csdds4ns8q85jnm0k4eclklti7p7a45@4ax.com...> I may have discovered a weird thermal runaway mechanism. Possibly as > temperature increases, more current shifts from the schottky junction > into the guard ring, reverse recovery dissipation increases, and boom.Recovery tempco is brutal. It's always a runaway (bistable) system. It's only a matter of having enough cooling / little enough dissipation that it doesn't quite get there.> I might try some silicon carbide. More Vf but no guard ring. >Guard ring is still there, they just don't always show it. Infineon datasheets I think usually go there, you see a step change around 4-6V forward drop, depending on temp. No good for surge (they're probably not made with big enough dies / heavy enough metallization to handle useful surge in the first place), but you have all that clean, juicy SiC performance at currents below there. I'm sad no one makes an 0.5A rated diode in SMA package. Tim -- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website: https://www.seventransistorlabs.com/
Reply by ●October 16, 20182018-10-16
On Tue, 16 Oct 2018 17:09:23 -0500, "Tim Williams" <tiwill@seventransistorlabs.com> wrote:>"John Larkin" <jjlarkin@highland_snip_technology.com> wrote in message >news:7r7csdds4ns8q85jnm0k4eclklti7p7a45@4ax.com... >> I may have discovered a weird thermal runaway mechanism. Possibly as >> temperature increases, more current shifts from the schottky junction >> into the guard ring, reverse recovery dissipation increases, and boom. > >Recovery tempco is brutal. It's always a runaway (bistable) system. It's >only a matter of having enough cooling / little enough dissipation that it >doesn't quite get there. > > >> I might try some silicon carbide. More Vf but no guard ring. >> > >Guard ring is still there, they just don't always show it. Infineon >datasheets I think usually go there, you see a step change around 4-6V >forward drop, depending on temp. No good for surge (they're probably not >made with big enough dies / heavy enough metallization to handle useful >surge in the first place), but you have all that clean, juicy SiC >performance at currents below there. > >I'm sad no one makes an 0.5A rated diode in SMA package. > >TimI'll be pushing maybe 10 to 15 amps in a short pulse, and I'll be back-biasing the diode fast at the end. I will be expecting numbers like 1.5 volts forward, which sounds safe based on your 4-6 volts for the guard ring. Most of the SiC parts are high voltage, 600 to 1200 roughly, and 2 to 20 amps or so. This is cute, but the price is insane. https://www.digikey.com/product-detail/en/genesic-semiconductor/GAP3SLT33-214/1242-1172-2-ND/3904847 There are some 1 amp parts in SMB. Maybe some melfs? -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Reply by ●October 17, 20182018-10-17
"John Larkin" <jjlarkin@highland_snip_technology.com> wrote in message news:jtqcsdde1hfmdg6odpreilrfdjks11kme9@4ax.com...> I'll be pushing maybe 10 to 15 amps in a short pulse, and I'll be > back-biasing the diode fast at the end. I will be expecting numbers > like 1.5 volts forward, which sounds safe based on your 4-6 volts for > the guard ring.To get there, you'll need pretty much a 4-8A diode. You'll have to check if it's worth the extra capacitance, or if you need to eat the losses with a bigger heatsink. There's always your old GaN friend, a few EPCxxx stacked would do almost as well, with a tiny fraction of the capacitance. And you already have some familiarity with mounting and sinking them. There are bigger, hotter (~600V) GaNs on the market now, which may be a better fit (less Vf than a stack).> This is cute, but the price is insane.^ ^ ^> > https://www.digikey.com/product-detail/en/genesic-semiconductor/GAP3SLT33-214/1242-1172-2-ND/3904847 >Yeah, like I said, they don't exist. :^) Tim -- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website: https://www.seventransistorlabs.com/
