>>Yeah, it's like doing chip design circa 1975--the NPNs are 100 times >>faster than the PNPs. :(>[snip]>That's stretching history just a tweak... PNP's back then were >_laterals_ ... 1MHz GBW product, if you were lucky.Right, and the NPNs were around 100 MHz. The fastest NPNs I have in my drawer are about 80 GHz, and the fastest PNPs other than the BFT92 are around 600 MHz. So we're back to a 100:1 ratio. Just today I replaced a BFT92 in a customer design with a 2N3906ish cascode pair running at several times the current. Barstids. Phil Hobbs
Well, it happened--the last fast PNP is EOL
Started by ●June 4, 2018
Reply by ●June 5, 20182018-06-05
Reply by ●June 5, 20182018-06-05
On Tue, 05 Jun 2018 12:48:16 -0700, Joerg <news@analogconsultants.com> wrote:>On 2018-06-05 12:27, Jim Thompson wrote: >> On Tue, 05 Jun 2018 12:22:47 -0700, Joerg <news@analogconsultants.com> >> wrote: >> >>> On 2018-06-05 12:19, Jim Thompson wrote: >>>> On Tue, 5 Jun 2018 18:38:27 +0000 (UTC), Uwe Bonnes >>>> <bon@hertz.tu-darmstadt.de> wrote: >>>> >>>>> John Larkin <jjlarkin@highlandtechnology.com> wrote: >>>>> .. >>>>>> I emailed ON and suggested they do some fast PNPs. They replied that >>>>>> demand is too small. >>>>> >>>>> I wondered that the military sector kept calm about the obsolation... >>>> >>>> Most of my chip designs, transferred to Lansdale, are on the military >>>> preferred components list... they seem to love unconditionally stable >>>> OpAmps, even if the open-loop gain is only 75dB... though I'm sure the >>>> sliding-class-A outputs and 10V/us slew-rates grab their attention as >>>> well ;-) >>>> >>> >>> Any radhard opamps in there with CM input range to the negative rail and >>> output to neg rail with pull-down? Will need that soon. >> >> Not from my just-out-of-school era. >> >> Today, I can easily design that for you... processing by X-Fab. >> > >I'll find some off-the-shelf from "the usual suspects", was just >wondering if there was a genuine Thompson amp around in rad-hard. > >Sometimes the just out of school designs are like VW Beetles. I have one >that runs off the conveyor belt since the 90's, now in Shenzhen, no end >in sight.The rabbit just keeps going... and going... and going... and going ;-) ...Jim Thompson -- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | STV, Queen Creek, AZ 85142 Skype: skypeanalog | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | Thinking outside the box... producing elegant solutions, by understanding what nature is hiding. "It is not in doing what you like, but in liking what you do that is the secret of happiness." -James Barrie
Reply by ●June 5, 20182018-06-05
On Tue, 05 Jun 2018 12:24:56 -0700, Jim Thompson <To-Email-Use-The-Envelope-Icon@On-My-Web-Site.com> wrote:>On Tue, 5 Jun 2018 15:07:16 -0400, Phil Hobbs ><pcdhSpamMeSenseless@electrooptical.net> wrote: > >[snip] >>> >>> The EOL of discretes does keep fast circuit design interesting. GaN >>> parts are cheap and fast, but don't come in P-channel! >>> >>> >> >>Yeah, it's like doing chip design circa 1975--the NPNs are 100 times >>faster than the PNPs. :( >> >[snip] > >That's stretching history just a tweak... PNP's back then were >_laterals_ ... 1MHz GBW product, if you were lucky.Point contact, Ft's measured in KHz. I think the first Touch-Tone phones used point contact transistors. Later parts were alloy junctions, then grown junctions. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Reply by ●June 5, 20182018-06-05
On Tue, 5 Jun 2018 14:15:19 -0700 (PDT), pcdhobbs@gmail.com wrote:>>>Yeah, it's like doing chip design circa 1975--the NPNs are 100 times >>>faster than the PNPs. :( > >>[snip] > >>That's stretching history just a tweak... PNP's back then were >>_laterals_ ... 1MHz GBW product, if you were lucky. > >Right, and the NPNs were around 100 MHz. The fastest NPNs I have in my drawer are about 80 GHz, and the fastest PNPs other than the BFT92 are around 600 MHz. So we're back to a 100:1 ratio. > >Just today I replaced a BFT92 in a customer design with a 2N3906ish cascode pair running at several times the current. > >Barstids. > >Phil HobbsShould be only about 3:1, based on mobility. Wonder what the gotcha is? Or is it that SiGe is only NPN? ...Jim Thompson -- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | STV, Queen Creek, AZ 85142 Skype: skypeanalog | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | Thinking outside the box... producing elegant solutions, by understanding what nature is hiding. "It is not in doing what you like, but in liking what you do that is the secret of happiness." -James Barrie
Reply by ●June 5, 20182018-06-05
On Tue, 05 Jun 2018 13:09:16 -0700, John Larkin <jjlarkin@highland_snip_technology.com> wrote:>On Tue, 5 Jun 2018 15:07:16 -0400, Phil Hobbs ><pcdhSpamMeSenseless@electrooptical.net> wrote: > >>On 06/05/2018 01:29 PM, John Larkin wrote: >>> On Tue, 5 Jun 2018 10:56:25 -0400, Phil Hobbs >>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>> >>>> On 06/04/2018 08:40 PM, Jim Thompson wrote: >>>>> On Mon, 4 Jun 2018 14:49:00 -0700 (PDT), Lasse Langwadt Christensen >>>>> <langwadt@fonz.dk> wrote: >>>>> >>>>>> mandag den 4. juni 2018 kl. 23.42.01 UTC+2 skrev Jim Thompson: >>>>>>> On 4 Jun 2018 14:04:31 -0700, Winfield Hill <hill@rowland.harvard.edu> >>>>>>> wrote: >>>>>>> >>>>>>>> bitrex wrote... >>>>>>>>> >>>>>>>>> >>>>>>>>> Should start a s.e.d. fabless semiconductor company and get them made >>>>>>>>> again, China will make whatever you like. You could advertise it as >>>>>>>>> exactly that "The Last Fast PNP" like the Last of the Mohicans or something. >>>>>>>> >>>>>>>> It's be nice to have it available in a SOT-323 SC70 package. >>>>>>> >>>>>>> Maybe you and Hobbs should buy a wafer or two? Then, as time moves >>>>>>> on, package them to suit the era? >>>>>>> >>>>>> >>>>>> digikey has 18000 in stock, $2,835 for 15000 how many wafers can you get for that? >>>>>> >>>>> >>>>> Discrete device wafers are dirt-cheap and low profitability. That's >>>>> why the devices are being phased out. Was anyone besides Hobbs buying >>>>> them? >>>>> >>>>> ...Jim Thompson >>>>> >>>> >>>> They sell for a lot more than digital transistors or BCX71s, and can't >>>> be much harder to make. Nexperia didn't take NXP's rf transistors or >>>> JFETs, so they're getting rid of them. >>>> >>>> Cheers >>>> >>>> Phil Hobbs >>> >>> The EOL of discretes does keep fast circuit design interesting. GaN >>> parts are cheap and fast, but don't come in P-channel! >>> >>> >> >>Yeah, it's like doing chip design circa 1975--the NPNs are 100 times >>faster than the PNPs. :( >> >>Cheers >> >>Phil Hobbs > >The PNP CK722 was the fastest transistor when I was a kid.CK760/761... my father was a Raytheon wholesaler... that's what twisted me away from architecture into electronics... toobs just didn't have any sex appeal ;-) ...Jim Thompson -- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | STV, Queen Creek, AZ 85142 Skype: skypeanalog | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | Thinking outside the box... producing elegant solutions, by understanding what nature is hiding. "It is not in doing what you like, but in liking what you do that is the secret of happiness." -James Barrie
Reply by ●June 5, 20182018-06-05
On Tue, 5 Jun 2018 14:15:19 -0700 (PDT), pcdhobbs@gmail.com wrote:>>>Yeah, it's like doing chip design circa 1975--the NPNs are 100 times >>>faster than the PNPs. :( > >>[snip] > >>That's stretching history just a tweak... PNP's back then were >>_laterals_ ... 1MHz GBW product, if you were lucky. > >Right, and the NPNs were around 100 MHz. The fastest NPNs I have in my drawer are about 80 GHz, and the fastest PNPs other than the BFT92 are around 600 MHz. So we're back to a 100:1 ratio. > >Just today I replaced a BFT92 in a customer design with a 2N3906ish cascode pair running at several times the current. > >Barstids. > >Phil HobbsON has some PNPs up to 850 MHz. Medium fast. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Reply by ●June 5, 20182018-06-05
On Tue, 05 Jun 2018 14:22:45 -0700, Jim Thompson <To-Email-Use-The-Envelope-Icon@On-My-Web-Site.com> wrote:>On Tue, 05 Jun 2018 13:09:16 -0700, John Larkin ><jjlarkin@highland_snip_technology.com> wrote: > >>On Tue, 5 Jun 2018 15:07:16 -0400, Phil Hobbs >><pcdhSpamMeSenseless@electrooptical.net> wrote: >> >>>On 06/05/2018 01:29 PM, John Larkin wrote: >>>> On Tue, 5 Jun 2018 10:56:25 -0400, Phil Hobbs >>>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>>> >>>>> On 06/04/2018 08:40 PM, Jim Thompson wrote: >>>>>> On Mon, 4 Jun 2018 14:49:00 -0700 (PDT), Lasse Langwadt Christensen >>>>>> <langwadt@fonz.dk> wrote: >>>>>> >>>>>>> mandag den 4. juni 2018 kl. 23.42.01 UTC+2 skrev Jim Thompson: >>>>>>>> On 4 Jun 2018 14:04:31 -0700, Winfield Hill <hill@rowland.harvard.edu> >>>>>>>> wrote: >>>>>>>> >>>>>>>>> bitrex wrote... >>>>>>>>>> >>>>>>>>>> >>>>>>>>>> Should start a s.e.d. fabless semiconductor company and get them made >>>>>>>>>> again, China will make whatever you like. You could advertise it as >>>>>>>>>> exactly that "The Last Fast PNP" like the Last of the Mohicans or something. >>>>>>>>> >>>>>>>>> It's be nice to have it available in a SOT-323 SC70 package. >>>>>>>> >>>>>>>> Maybe you and Hobbs should buy a wafer or two? Then, as time moves >>>>>>>> on, package them to suit the era? >>>>>>>> >>>>>>> >>>>>>> digikey has 18000 in stock, $2,835 for 15000 how many wafers can you get for that? >>>>>>> >>>>>> >>>>>> Discrete device wafers are dirt-cheap and low profitability. That's >>>>>> why the devices are being phased out. Was anyone besides Hobbs buying >>>>>> them? >>>>>> >>>>>> ...Jim Thompson >>>>>> >>>>> >>>>> They sell for a lot more than digital transistors or BCX71s, and can't >>>>> be much harder to make. Nexperia didn't take NXP's rf transistors or >>>>> JFETs, so they're getting rid of them. >>>>> >>>>> Cheers >>>>> >>>>> Phil Hobbs >>>> >>>> The EOL of discretes does keep fast circuit design interesting. GaN >>>> parts are cheap and fast, but don't come in P-channel! >>>> >>>> >>> >>>Yeah, it's like doing chip design circa 1975--the NPNs are 100 times >>>faster than the PNPs. :( >>> >>>Cheers >>> >>>Phil Hobbs >> >>The PNP CK722 was the fastest transistor when I was a kid. > >CK760/761... my father was a Raytheon wholesaler... that's what >twisted me away from architecture into electronics... toobs just >didn't have any sex appeal ;-) > > ...Jim ThompsonI liked tubes, partly because I got them free. The first thing I designed professionally was a radiation counter that used those circular-readout gas discharge tubes, Decatrons or something. I was still in high school. Exotic tubes were fun. Klystrons, thyratrons, PMTs, CRTs, intensifiers, sniperscope imagers, flashtubes, acorns, high voltage, weird RF jugs, like that. I wanted an xray tube but never got one. I had an interview with a priggish guy; I told him that I liked tubes because they were harder to blow up than transistors. He sniffed "that won't do" and dismissed me. I said the same thing to the next guy and he laughed and hired me. I designed over $100 million worth of stuff for him. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Reply by ●June 5, 20182018-06-05
Am 05.06.2018 um 22:25 schrieb Joerg:> On 2018-06-05 13:07, Gerhard Hoffmann wrote: >> Am 05.06.2018 um 21:22 schrieb Joerg:>>> Any radhard opamps in there with CM input range to the negative rail >>> and output to neg rail with pull-down? Will need that soon. >>> >> We did use lots of OP484 :-) >> > > Though I am sure the taxpayer was ultimately on the hook for those :-)You must get used to a new set of cost metrics that is completely different to that you are used to. If it is on the qualified parts list, and it is available in less than a year, choose it. You will probably build some engineering samples and one flight sample. For the engineering samples, you can use commercial parts. When a commercial quad op amp costs $5, nobody cares. Just deciding to order ten of them costs more. (Just kidding about you building the flight sample. I was allowed to touch it to bring it up, but there was no soldering iron in the clean room. They had quality controlled solder ladies for that.) We had nice part decals that could accept both plastic SO-14 and the ceramic flat packs for flight. It is important that this is done according to flight board design rules, so you do not have to re-do the board layout. The cost of qualifying a new part will dwarf the part's commercial price. Irradiation will cost a few grand, that's no problem. But you must find someone to do the paperwork. That generates work for someone higher in the hierarchy than you, and that is usually frowned upon. I found it impossible to get a fast-ish JFET op amp for my 1:3000 pulse width stretcher. I finally gave up and designed around the problem with discrete FETs. That exploded the area of the stretcher :-( and it was ugly. Then they found they had leftover JFETs from a previous mission, _please_ , can you use this type? Aaaarghhh, JFETs are all individuals, even if they would not be of a different type. Nobody will complain if you want 5 inductors of 470nH each. But if you want 1 each of 330n, 390n, 470n, 560n... hell will break loose. But you may not know in advance what your flight crystal looks like electrically, and the inductance needed to pull it to the right center frequency may vary. Cheers, Gerhard (ich bin jetzt ensthaft unterhopft!)
Reply by ●June 5, 20182018-06-05
On 2018-06-05 16:11, Gerhard Hoffmann wrote:> Am 05.06.2018 um 22:25 schrieb Joerg: >> On 2018-06-05 13:07, Gerhard Hoffmann wrote: >>> Am 05.06.2018 um 21:22 schrieb Joerg: > >>>> Any radhard opamps in there with CM input range to the negative rail >>>> and output to neg rail with pull-down? Will need that soon. >>>> >>> We did use lots of OP484 :-) >>> >> >> Though I am sure the taxpayer was ultimately on the hook for those :-) > > You must get used to a new set of cost metrics that is completely > different to that you are used to. > > If it is on the qualified parts list, and it is available in less > than a year, choose it. You will probably build some engineering > samples and one flight sample. For the engineering samples, you can > use commercial parts. When a commercial quad op amp costs $5, > nobody cares. Just deciding to order ten of them costs more. > > (Just kidding about you building the flight sample. I was allowed to > touch it to bring it up, but there was no soldering iron in the clean > room. They had quality controlled solder ladies for that.) >Yup, I am a bit familiar with that market. Now where it is going towards the private sector cost begins to matter.> We had nice part decals that could accept both plastic SO-14 and the > ceramic flat packs for flight. It is important that this is done > according to flight board design rules, so you do not have to re-do > the board layout. > > The cost of qualifying a new part will dwarf the part's commercial > price. Irradiation will cost a few grand, that's no problem. But you > must find someone to do the paperwork. That generates work for someone > higher in the hierarchy than you, and that is usually frowned upon. >Don't spoil someone's golf game by generating a lot of work 8-)> I found it impossible to get a fast-ish JFET op amp for my 1:3000 > pulse width stretcher. I finally gave up and designed around the > problem with discrete FETs. That exploded the area of the stretcher :-( > and it was ugly. Then they found they had leftover JFETs from a > previous mission, _please_ , can you use this type? Aaaarghhh, JFETs are > all individuals, even if they would not be of a different type. > > Nobody will complain if you want 5 inductors of 470nH each. > But if you want 1 each of 330n, 390n, 470n, 560n... hell will break > loose. But you may not know in advance what your flight crystal > looks like electrically, and the inductance needed to pull it > to the right center frequency may vary. >My stuff is all in the lower frequency domain, what you guys would consider jittery DC.> Cheers, Gerhard > > (ich bin jetzt ensthaft unterhopft!) >Come over here, I can help. I've got a Pliny the Elder clone in bottles (not many left though) and another will be dry-hopped this week, then again early next week. The Belgian Quadrupel also looks promising and clocks in above 9% but that has to rest in the secondary fermenter and then in bottles for another two months. Here is a Belgian Tripel blowing off its Kraeusen: http://www.analogconsultants.com/ng/brew/Blowoff.JPG -- Regards, Joerg http://www.analogconsultants.com/
Reply by ●June 5, 20182018-06-05
On 06/05/18 17:21, Jim Thompson wrote:> On Tue, 5 Jun 2018 14:15:19 -0700 (PDT), pcdhobbs@gmail.com wrote: > >>>> Yeah, it's like doing chip design circa 1975--the NPNs are 100 times >>>> faster than the PNPs. :( >> >>> [snip] >> >>> That's stretching history just a tweak... PNP's back then were >>> _laterals_ ... 1MHz GBW product, if you were lucky. >> >> Right, and the NPNs were around 100 MHz. The fastest NPNs I have in my drawer are about 80 GHz, and the fastest PNPs other than the BFT92 are around 600 MHz. So we're back to a 100:1 ratio. >> >> Just today I replaced a BFT92 in a customer design with a 2N3906ish cascode pair running at several times the current. >> >> Barstids. >> >> Phil Hobbs > > Should be only about 3:1, based on mobility. Wonder what the gotcha > is? Or is it that SiGe is only NPN?Dunno. Germanium hole mobility is around 2000 cm**2/V/s, which is higher than electron mobility in silicon, and dramatically higher than silicon hole mobility. I suspect that nobody expected people to design complementary circuits that fast. I do a certain amount of gigahertz stuff, but lots of the time I'm using superfast devices in unusual ways, e.g. my fave pHEMT/SiGe NPN cascode, which I'd claim is the best wideband front end building block out there. BTW CEL have a bunch of newish pHEMTs to replace the late lamented NE3509 etc. 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
