On 02/13/2013 04:50 PM, tm wrote:> > "Phil Hobbs" <pcdhSpamMeSenseless@electrooptical.net> wrote in message > news:fZudnZV219rrmoHMnZ2dnUVZ_tudnZ2d@supernews.com... >> On 02/13/2013 03:38 PM, tm wrote: >>> >>> Hi Phil, >>> >>> Do you have the manual for the 72BD? >>> >>> regards >> >> I have a softcopy of the user's manual, but no service manual or >> schematics. >> >> Cheers >> >> Phil Hobbs >> >> -- > > If I find an electronic copy I'll let you know. > > tmThanks! 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
Curve tracers
Started by ●February 11, 2013
Reply by ●February 13, 20132013-02-13
Reply by ●February 13, 20132013-02-13
On Wed, 13 Feb 2013 10:56:04 -0500, Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:>On 2/13/2013 12:35 AM, josephkk wrote: >> On Mon, 11 Feb 2013 17:25:17 -0500, Phil Hobbs >> <pcdhSpamMeSenseless@electrooptical.net> wrote: >> >>> >>>> Hey, doesn't one of your kids want to get his hands dirty with some >>>> software work? If you charge the client a lower rate for that work =he>>>> could still sock away a lot of money towards his first souped-up =TransAm>>>> with mag wheels and all that :-) >>>> >>> >>> Brr. My son is _not_ "going to drive me to drinkin'" if I can help =it,>>> but anyway that wouldn't be his style. He's busy writing Arm Cortex =M3>>> firmware at the moment, but since he's learning C at the same time, =he>>> isn't fast at it yet. (He did get a task switcher working, using a >>> linked list of tasks handled by the systick interrupt handler.) >> >> Hey it may be a bit basic, but it works. It is an adequate approach =for>> very simple rate monotonic schedulers but the linked list time will =start>> to hurt if things get complicated and crowded schedule wise. He needs=to>> separate the switcher from the task manager. Then schedule time for =the>> task manger as a task. >>> > >It isn't really a multitasking OS, just a way of scheduling stuff for=20 >future reference, e.g. looking after a delta-sigma ADC that needs a=20 >3-million clock reset interval. For now the list elements consist of a=20 >time stamp, a pointer to a flag variable, and the tail pointer. The=20 >systick interrupt just looks at the next thing in the list to see if the==20>time stamp matches, and if so, pops the element, sets the flag, and=20 >checks the next one in case it has the same time stamp. > >All the work is done by functions called from the housekeeping loop.=20 >None of them takes enough time that it needs to be preempted--it's quick==20>stuff like setting up a DMA to redraw the screen, taking a data point,=20 >sending a character to the USB serial port, that sort of stuff. (I've=20 >been doing embedded stuff off and on for a dozen or so years, but I'm=20 >nobody's idea of an expert at it.) > >Cheers > >Phil HobbsIf he really goes into software he will eventually take an os course. If he does this now he will be miles ahead by then. I know, this was part = of my first os course. He could also delve into searching and sorting, b-trees, and timer queues. All relevant at the os level. Serching, sorting and b-trees (and the variants) are all generally applicable. ?-)
Reply by ●February 13, 20132013-02-13
On Wed, 13 Feb 2013 11:16:46 -0800, John Larkin <jlarkin@highlandtechnology.com> wrote:>On Wed, 13 Feb 2013 13:41:34 -0500, Phil Hobbs ><pcdhSpamMeSenseless@electrooptical.net> wrote: > >>On 2/13/2013 12:27 PM, John Larkin wrote: >>> On Wed, 13 Feb 2013 11:34:36 -0500, Phil Hobbs >>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>> >>>> On 2/13/2013 11:30 AM, John Larkin wrote: >>>>> On Wed, 13 Feb 2013 11:01:28 -0500, Phil Hobbs >>>>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>>>> >>>>>> On 2/13/2013 10:24 AM, John Larkin wrote: >>>>>>> On Tue, 12 Feb 2013 21:57:13 -0800, josephkk =<joseph_barrett@sbcglobal.net>>>>>>>> wrote: >>>>>>> >>>>>>>> On Tue, 12 Feb 2013 11:59:36 -0500, Phil Hobbs >>>>>>>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>>>>>>> >>>>>>>>> On 02/11/2013 05:29 PM, mike wrote: >>>>>>>>>> On 2/11/2013 10:42 AM, Phil Hobbs wrote: >>>>>>>>>>> Hi, all--I'm looking for a curve tracer. Ideally it would be >>>>>>>>>>> USB-powered, and come with software that would make plots and=generate>>>>>>>>>>> data files. Nothing too fancy, just I_C and I_B vs V_CE and =V_BE, and>>>>>>>>>>> the equivalent for FETs. A plus would be automatic =computation of beta,>>>>>>>>>>> V_GSth, Early voltage, and even some of the DC SPICE =parameters like>>>>>>>>>>> emitter and base resistance. >>>>>>>>>>> >>>>>>>>>>> It would be especially nice if it could measure gate current =with a>>>>>>>>>>> resolution of a picoamp or less. >>>>>>>>>>> >>>>>>>>>>> Any suggestions? >>>>>>>>>>> >>>>>>>>>>> Cheers >>>>>>>>>>> >>>>>>>>>>> Phil Hobbs >>>>>>>>>>> >>>>>>>>>> You're asking a LOT. >>>>>>>>>> >>>>>>>>>> A curve tracer is a qualitative measurement device. >>>>>>>>>> I use a TEK 7CT1N for simple stuff and only drag out the 576 >>>>>>>>>> when I gotta have it. I use it more for blowing shorts >>>>>>>>>> out of NiCd's than for testing transistors. >>>>>>>>> >>>>>>>>> It's only qualitative because it only has a CRT readout. A =curve tracer>>>>>>>>> is a sort of stripped-down version of a semiconductor parameter=analyzer>>>>>>>>> such as an HP4145B, which is very quantitative indeed. (I =wouldn't mind>>>>>>>>> one of those--I used to have one at my PPOE, but it takes a lot=of rack>>>>>>>>> space.) >>>>>>>>>> >>>>>>>>>> If you try to make one using a series of fixed measurements >>>>>>>>>> and graphing them, you get mixed results. I tried that back in=the day.>>>>>>>>>> The device thermal time constants significantly affect the =measurement>>>>>>>>>> details. >>>>>>>>> >>>>>>>>> Yup. You have to be quick about it if you're dissipating any >>>>>>>>> significant power. Of course with a MCU and DACs, you can do =the>>>>>>>>> measurements out of order to keep the average dissipation =constant.>>>>>>>>> >>>>>>>>>> >>>>>>>>>> And the parasitics of the test fixture swamp device =parasitics, unless you>>>>>>>>>> use network analyzer fixturing. >>>>>>>>> >>>>>>>>> Not necessarily. My Boonton 72BD easily measures a few =femtofarads.>>>>>>>>> You just null it out before attaching the hot lead to the =device.>>>>>>>>> >>>>>>>>>> >>>>>>>>>> Picoamps to amps is a wide dynamic range and will require some=serious>>>>>>>>>> attention to range switching. >>>>>>>>>> >>>>>>>>>> Not saying it can't be done, just that it's not nearly as =simple as it>>>>>>>>>> sounds. >>>>>>>>> >>>>>>>>> I don't need amps--I'd be quite happy with a 50-100 mA top end.=Most of>>>>>>>>> what I need a curve tracer for is measuring weird RF devices =with lousy>>>>>>>>> datasheets. >>>>>>>>> >>>>>>>>> Cheers >>>>>>>>> >>>>>>>>> Phil Hobbs >>>>>>>> >>>>>>>> Oh foo, the fixturing for RF devices can get really weird and =expensive.>>>>>>>> Those will be on you if we get together and create a USB curve =tracer.>>>>>>>> >>>>>>>> ?-) >>>>>>> >>>>>>> The problem with a lot of (most) RF devices is their terrible DC >>>>>>> characterization. Like, no DC curves at all. The s-params give =you a rough idea>>>>>>> of the high-speed behavior. >>>>>>> >>>>>>> RF data sheets often say to just adjust the bias until it works. =The specified>>>>>>> breakdown voltage specs are not useful, because sometimes they =assume, without>>>>>>> saying so, that the drain voltage can swing to multiples of the =supply voltage,>>>>>>> and factor that in. Parts spec'd at 6 volts can cheerfully =avalanche at 25. A>>>>>>> detector diode might be spec'd a 2 volts reverse, woth no curves =at all, because>>>>>>> they figure that's all a detector user needs to know. >>>>>> >>>>>> I'm normally working down on the low frequency end anyway, e.g. a =100>>>>>> MHz biochip front end using 14-GHz pHEMTs and 40-GHz bipolars, so =simple>>>>>> concepts like C_JC and R_BB' still work OK. Boontons rule! >>>>> >>>>> Right. You have to make your own curves, like C-vs-V. That's why a =USB>>>>> curve tracer should measure capacitances, too. >>>>> >>>>> Heck, somebody here with a curve tracer and a Booonton could set up=a>>>>> garage operation to characterize parts for a fee. Modelithics does >>>>> that for big bucks. >>>>> >>>>> >>>> >>>> Have you measured the leakage of the Boonton's biasing system? =Maybe>>>> one could just hang a little board on the back of one, and do it =that way.>>>> >>>> Cheers >>>> >>>> Phil Hobbs >>> >>> Good idea. Design the USB curve tracer to run through a Boonton 72 to >>> get the capacitances. The Boonton has an analog output on the back. >>> But it would be more fun to do it ourselves. >>> >>> >> >>Sure, I'm in. Do you think it might actually sell? >> > >Well, some for sure. Ten a year? 200?=20 > >Time to start scribbling.We might make two or more models. Some dependency on price, features, = and performance for sales outside this ng. Within the ng we could probably make and deploy one to two dozen. Then with real world experience i think we could do a better job of selling it outside. ?-)
Reply by ●February 14, 20132013-02-14
On Wed, 13 Feb 2013 11:54:47 -0800, Jon Kirwan <jonk@infinitefactors.org> wrote:>On Wed, 13 Feb 2013 11:16:46 -0800, John Larkin ><jlarkin@highlandtechnology.com> wrote: > >><snip> >>Well, some for sure. Ten a year? 200?=20 >> >>Time to start scribbling. > >I'd be happy to volunteer embedded software work. Pretty much >what I do is measurement instrumentation software and >associated Windows software, where needed. > >For EM1 level: BF, BR, IS >Measure ambient temperature near the BJT [and hold Vbc near 0 >to avoid power dissipation as is assumed in EM1] and also >require doing this at several ambient temperatures to get Eg >and perhaps the power of T [nominally 3 in the IS(T) >equation, but if allowed to change freely then becomes part >of the curve fitting algorithm, which I can also write.] By >the way, the Tek 577 achieves base currents of 5nA with the >177 porch. Not sure how low you'd like to achieve. > >For EM2 level: Ohmic re', rc', and rb'; Cjeo, Cjco, >C-substrate; forward and reverse transit times; gradients for >the EB and CB capacitances and their barrier potentials. >Things aren't always straightword. If you attempt to measure >re', for example, by driving known base currents and >observing Vce (with open collector), there is a flyback >effect caused by a decrease of BR at low currents which >should be taken into account. The re' value is best gained as >1/slope near that flyback point than somewhere else on the >curve. But not confused by the actual flyback itself. > >JonEbbers-Moll is ok for bipolar but i would like to get Gummel-Poon models. MOS models are a different subject that i didn't get to study. ?-)
Reply by ●February 14, 20132013-02-14
On Wed, 13 Feb 2013 12:32:17 -0800, mike <ham789@netzero.net> wrote:>On 2/13/2013 9:36 AM, Fred Abse wrote: >> On Tue, 12 Feb 2013 15:30:25 -0800, mike wrote: >> >>> Methinks you're applying apples thinking to oranges. If you take a >>> measurement at a fixed operating point, whether it be DC or pulsed, =your>>> average junction temperature is likely to be different from what it =would>>> be in an analog curve tracer like the 576. All those measurement =points>>> are not likely to line up on exactly the same line you'd see with an >>> analog plot. Nothing wrong with that. If you want data at a =particular>>> operating point, that's the way to go. >>> If you want a curve tracer, like it says in the subject line, you may=not>>> see what you expect. >> >> The basic 576 *incorporates* pulsed measurements, as well as swept. >> 300 microseconds and 80 microseconds. Bottom row of buttons on the =right>> hand side. It does that at a single operating point at a time, if you =want>> a plot, you have to draw it by hand. > >You insist on talking about how a 576 works. >I'm talking about the measurement you're making. >If what you're measuring is what you want, you're good to go. > >If you want to DESIGN a curve tracer, you might want to be >concerned with the relationship between what you measure >and the picture on the display. >For a qualitative display, it doesn't matter a lot. > >If you're reading out three digits of precision, you need to think about==20>dynamic temperature effects. > >You can do it either way...just understand the difference. >> >> Methinks you are not that familiar with the 576. >> I use one or other of the two in my lab on an almost daily basis. > >You have me beat. I have only one. I bow to your quantitative >superiority. All I can tell you is that, back in the day, I tried to >build a curve tracer that graphed static measurements. >I didn't like the result and aborted the project. It was not as trivial >as I thought.BTDT. I have some grasp and this project looks like a ton of fun but likely beyond the best i can do alone, even if i had the time. To avoid the half ton power supplies we use pulsed mode measurements. With so = much of everything now being focused on low power for battery use, USB powered could do much of what we want, and add a brick for moderate power pulsed measurements. Good clean ramps and steps (in both voltage and current = are not simple) and adding pulse measurements to that gets seriously hard core. Some here could do it all by themselves, but i suspect no one = would volunteer to do it all without a contract.>> >> The 176 is an addon unit that fits in place of the standard test =fixture,>> that goes to 200 amps. It is pulsed mode only, using delay line charge >> storage, to avoid enormous, heavy power supplies, unlike the 175, that >> needed two people or a fork truck to carry it. >>
Reply by ●February 14, 20132013-02-14
On Wed, 13 Feb 2013 14:29:35 -0800, Fred Abse <excretatauris@invalid.invalid> wrote:>On Wed, 13 Feb 2013 12:32:17 -0800, mike wrote: > >> If you're reading out three digits of precision, you need to think =about>> dynamic temperature effects. > >If you're reading three digits of precision, I would recommend reading =the>sheaf of application notes put out by Keithley on the subject of >semiconductor characterization. It's decidedly nontrivial.Thanks for the heads-up. I am not sure we are going to do anything that need these, but it will be nice to have them. ?-)
Reply by ●February 14, 20132013-02-14
On Wed, 13 Feb 2013 20:01:26 -0800, josephkk <joseph_barrett@sbcglobal.net> wrote:>On Wed, 13 Feb 2013 11:54:47 -0800, Jon Kirwan <jonk@infinitefactors.org> >wrote: > >>On Wed, 13 Feb 2013 11:16:46 -0800, John Larkin >><jlarkin@highlandtechnology.com> wrote: >> >>><snip> >>>Well, some for sure. Ten a year? 200? >>> >>>Time to start scribbling. >> >>I'd be happy to volunteer embedded software work. Pretty much >>what I do is measurement instrumentation software and >>associated Windows software, where needed. >> >>For EM1 level: BF, BR, IS >>Measure ambient temperature near the BJT [and hold Vbc near 0 >>to avoid power dissipation as is assumed in EM1] and also >>require doing this at several ambient temperatures to get Eg >>and perhaps the power of T [nominally 3 in the IS(T) >>equation, but if allowed to change freely then becomes part >>of the curve fitting algorithm, which I can also write.] By >>the way, the Tek 577 achieves base currents of 5nA with the >>177 porch. Not sure how low you'd like to achieve. >> >>For EM2 level: Ohmic re', rc', and rb'; Cjeo, Cjco, >>C-substrate; forward and reverse transit times; gradients for >>the EB and CB capacitances and their barrier potentials. >>Things aren't always straightword. If you attempt to measure >>re', for example, by driving known base currents and >>observing Vce (with open collector), there is a flyback >>effect caused by a decrease of BR at low currents which >>should be taken into account. The re' value is best gained as >>1/slope near that flyback point than somewhere else on the >>curve. But not confused by the actual flyback itself. >> >>Jon > >Ebbers-Moll is ok for bipolar but i would like to get Gummel-Poon models.I agree. I was just pointing out that it starts out interesting at the DC level and gets even more interesting later. This is an area I really enjoy. Take VA. The easiest way is to extrapolate Ic vs Vce in the CE configuration. But the error can be very high. Another way is to measure the slope very near Vbc=0. Another way, though, is using ln(Ic) vs Vbe at two different Vbc values, computing it as a ratio. And it's better done that way. But that assumes basewidth modulation is negligible. So additional tests using different Vbe to see how well the assumption holds may be useful. I'll be picking up a few BJT testers; 575, 577, and fairchild 6200, in a couple of months.>MOS models are a different subject that i didn't get to study.I'm in a similar boat. I've only basic knowledge here. They don't seem as intriguing, though. Jon
Reply by ●February 14, 20132013-02-14
On Wed, 13 Feb 2013 19:56:34 -0800, the renowned josephkk <joseph_barrett@sbcglobal.net> wrote:>On Wed, 13 Feb 2013 11:16:46 -0800, John Larkin ><jlarkin@highlandtechnology.com> wrote: > >>On Wed, 13 Feb 2013 13:41:34 -0500, Phil Hobbs >><pcdhSpamMeSenseless@electrooptical.net> wrote: >> >>>On 2/13/2013 12:27 PM, John Larkin wrote: >>>> On Wed, 13 Feb 2013 11:34:36 -0500, Phil Hobbs >>>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>>> >>>>> On 2/13/2013 11:30 AM, John Larkin wrote: >>>>>> On Wed, 13 Feb 2013 11:01:28 -0500, Phil Hobbs >>>>>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>>>>> >>>>>>> On 2/13/2013 10:24 AM, John Larkin wrote: >>>>>>>> On Tue, 12 Feb 2013 21:57:13 -0800, josephkk <joseph_barrett@sbcglobal.net> >>>>>>>> wrote: >>>>>>>> >>>>>>>>> On Tue, 12 Feb 2013 11:59:36 -0500, Phil Hobbs >>>>>>>>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>>>>>>>> >>>>>>>>>> On 02/11/2013 05:29 PM, mike wrote: >>>>>>>>>>> On 2/11/2013 10:42 AM, Phil Hobbs wrote: >>>>>>>>>>>> Hi, all--I'm looking for a curve tracer. Ideally it would be >>>>>>>>>>>> USB-powered, and come with software that would make plots and generate >>>>>>>>>>>> data files. Nothing too fancy, just I_C and I_B vs V_CE and V_BE, and >>>>>>>>>>>> the equivalent for FETs. A plus would be automatic computation of beta, >>>>>>>>>>>> V_GSth, Early voltage, and even some of the DC SPICE parameters like >>>>>>>>>>>> emitter and base resistance. >>>>>>>>>>>> >>>>>>>>>>>> It would be especially nice if it could measure gate current with a >>>>>>>>>>>> resolution of a picoamp or less. >>>>>>>>>>>> >>>>>>>>>>>> Any suggestions? >>>>>>>>>>>> >>>>>>>>>>>> Cheers >>>>>>>>>>>> >>>>>>>>>>>> Phil Hobbs >>>>>>>>>>>> >>>>>>>>>>> You're asking a LOT. >>>>>>>>>>> >>>>>>>>>>> A curve tracer is a qualitative measurement device. >>>>>>>>>>> I use a TEK 7CT1N for simple stuff and only drag out the 576 >>>>>>>>>>> when I gotta have it. I use it more for blowing shorts >>>>>>>>>>> out of NiCd's than for testing transistors. >>>>>>>>>> >>>>>>>>>> It's only qualitative because it only has a CRT readout. A curve tracer >>>>>>>>>> is a sort of stripped-down version of a semiconductor parameter analyzer >>>>>>>>>> such as an HP4145B, which is very quantitative indeed. (I wouldn't mind >>>>>>>>>> one of those--I used to have one at my PPOE, but it takes a lot of rack >>>>>>>>>> space.) >>>>>>>>>>> >>>>>>>>>>> If you try to make one using a series of fixed measurements >>>>>>>>>>> and graphing them, you get mixed results. I tried that back in the day. >>>>>>>>>>> The device thermal time constants significantly affect the measurement >>>>>>>>>>> details. >>>>>>>>>> >>>>>>>>>> Yup. You have to be quick about it if you're dissipating any >>>>>>>>>> significant power. Of course with a MCU and DACs, you can do the >>>>>>>>>> measurements out of order to keep the average dissipation constant. >>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> And the parasitics of the test fixture swamp device parasitics, unless you >>>>>>>>>>> use network analyzer fixturing. >>>>>>>>>> >>>>>>>>>> Not necessarily. My Boonton 72BD easily measures a few femtofarads. >>>>>>>>>> You just null it out before attaching the hot lead to the device. >>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> Picoamps to amps is a wide dynamic range and will require some serious >>>>>>>>>>> attention to range switching. >>>>>>>>>>> >>>>>>>>>>> Not saying it can't be done, just that it's not nearly as simple as it >>>>>>>>>>> sounds. >>>>>>>>>> >>>>>>>>>> I don't need amps--I'd be quite happy with a 50-100 mA top end. Most of >>>>>>>>>> what I need a curve tracer for is measuring weird RF devices with lousy >>>>>>>>>> datasheets. >>>>>>>>>> >>>>>>>>>> Cheers >>>>>>>>>> >>>>>>>>>> Phil Hobbs >>>>>>>>> >>>>>>>>> Oh foo, the fixturing for RF devices can get really weird and expensive. >>>>>>>>> Those will be on you if we get together and create a USB curve tracer. >>>>>>>>> >>>>>>>>> ?-) >>>>>>>> >>>>>>>> The problem with a lot of (most) RF devices is their terrible DC >>>>>>>> characterization. Like, no DC curves at all. The s-params give you a rough idea >>>>>>>> of the high-speed behavior. >>>>>>>> >>>>>>>> RF data sheets often say to just adjust the bias until it works. The specified >>>>>>>> breakdown voltage specs are not useful, because sometimes they assume, without >>>>>>>> saying so, that the drain voltage can swing to multiples of the supply voltage, >>>>>>>> and factor that in. Parts spec'd at 6 volts can cheerfully avalanche at 25. A >>>>>>>> detector diode might be spec'd a 2 volts reverse, woth no curves at all, because >>>>>>>> they figure that's all a detector user needs to know. >>>>>>> >>>>>>> I'm normally working down on the low frequency end anyway, e.g. a 100 >>>>>>> MHz biochip front end using 14-GHz pHEMTs and 40-GHz bipolars, so simple >>>>>>> concepts like C_JC and R_BB' still work OK. Boontons rule! >>>>>> >>>>>> Right. You have to make your own curves, like C-vs-V. That's why a USB >>>>>> curve tracer should measure capacitances, too. >>>>>> >>>>>> Heck, somebody here with a curve tracer and a Booonton could set up a >>>>>> garage operation to characterize parts for a fee. Modelithics does >>>>>> that for big bucks. >>>>>> >>>>>> >>>>> >>>>> Have you measured the leakage of the Boonton's biasing system? Maybe >>>>> one could just hang a little board on the back of one, and do it that way. >>>>> >>>>> Cheers >>>>> >>>>> Phil Hobbs >>>> >>>> Good idea. Design the USB curve tracer to run through a Boonton 72 to >>>> get the capacitances. The Boonton has an analog output on the back. >>>> But it would be more fun to do it ourselves. >>>> >>>> >>> >>>Sure, I'm in. Do you think it might actually sell? >>> >> >>Well, some for sure. Ten a year? 200? >> >>Time to start scribbling. > >We might make two or more models. Some dependency on price, features, and >performance for sales outside this ng. Within the ng we could probably >make and deploy one to two dozen. Then with real world experience i >think we could do a better job of selling it outside. > >?-)How practical do you think it would it be to add temperature characterization? I'd be interested in that. Best regards, Spehro Pefhany -- "it's the network..." "The Journey is the reward" speff@interlog.com Info for manufacturers: http://www.trexon.com Embedded software/hardware/analog Info for designers: http://www.speff.com
Reply by ●February 14, 20132013-02-14
On 2/14/2013 8:04 AM, Spehro Pefhany wrote:> On Wed, 13 Feb 2013 19:56:34 -0800, the renowned josephkk > <joseph_barrett@sbcglobal.net> wrote: > >> On Wed, 13 Feb 2013 11:16:46 -0800, John Larkin >> <jlarkin@highlandtechnology.com> wrote: >> >>> On Wed, 13 Feb 2013 13:41:34 -0500, Phil Hobbs >>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>> >>>> On 2/13/2013 12:27 PM, John Larkin wrote: >>>>> On Wed, 13 Feb 2013 11:34:36 -0500, Phil Hobbs >>>>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>>>> >>>>>> On 2/13/2013 11:30 AM, John Larkin wrote: >>>>>>> On Wed, 13 Feb 2013 11:01:28 -0500, Phil Hobbs >>>>>>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>>>>>> >>>>>>>> On 2/13/2013 10:24 AM, John Larkin wrote: >>>>>>>>> On Tue, 12 Feb 2013 21:57:13 -0800, josephkk <joseph_barrett@sbcglobal.net> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>>> On Tue, 12 Feb 2013 11:59:36 -0500, Phil Hobbs >>>>>>>>>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>>>>>>>>> >>>>>>>>>>> On 02/11/2013 05:29 PM, mike wrote: >>>>>>>>>>>> On 2/11/2013 10:42 AM, Phil Hobbs wrote: >>>>>>>>>>>>> Hi, all--I'm looking for a curve tracer. Ideally it would be >>>>>>>>>>>>> USB-powered, and come with software that would make plots and generate >>>>>>>>>>>>> data files. Nothing too fancy, just I_C and I_B vs V_CE and V_BE, and >>>>>>>>>>>>> the equivalent for FETs. A plus would be automatic computation of beta, >>>>>>>>>>>>> V_GSth, Early voltage, and even some of the DC SPICE parameters like >>>>>>>>>>>>> emitter and base resistance. >>>>>>>>>>>>> >>>>>>>>>>>>> It would be especially nice if it could measure gate current with a >>>>>>>>>>>>> resolution of a picoamp or less. >>>>>>>>>>>>> >>>>>>>>>>>>> Any suggestions? >>>>>>>>>>>>> >>>>>>>>>>>>> Cheers >>>>>>>>>>>>> >>>>>>>>>>>>> Phil Hobbs >>>>>>>>>>>>> >>>>>>>>>>>> You're asking a LOT. >>>>>>>>>>>> >>>>>>>>>>>> A curve tracer is a qualitative measurement device. >>>>>>>>>>>> I use a TEK 7CT1N for simple stuff and only drag out the 576 >>>>>>>>>>>> when I gotta have it. I use it more for blowing shorts >>>>>>>>>>>> out of NiCd's than for testing transistors. >>>>>>>>>>> >>>>>>>>>>> It's only qualitative because it only has a CRT readout. A curve tracer >>>>>>>>>>> is a sort of stripped-down version of a semiconductor parameter analyzer >>>>>>>>>>> such as an HP4145B, which is very quantitative indeed. (I wouldn't mind >>>>>>>>>>> one of those--I used to have one at my PPOE, but it takes a lot of rack >>>>>>>>>>> space.) >>>>>>>>>>>> >>>>>>>>>>>> If you try to make one using a series of fixed measurements >>>>>>>>>>>> and graphing them, you get mixed results. I tried that back in the day. >>>>>>>>>>>> The device thermal time constants significantly affect the measurement >>>>>>>>>>>> details. >>>>>>>>>>> >>>>>>>>>>> Yup. You have to be quick about it if you're dissipating any >>>>>>>>>>> significant power. Of course with a MCU and DACs, you can do the >>>>>>>>>>> measurements out of order to keep the average dissipation constant. >>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> And the parasitics of the test fixture swamp device parasitics, unless you >>>>>>>>>>>> use network analyzer fixturing. >>>>>>>>>>> >>>>>>>>>>> Not necessarily. My Boonton 72BD easily measures a few femtofarads. >>>>>>>>>>> You just null it out before attaching the hot lead to the device. >>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> Picoamps to amps is a wide dynamic range and will require some serious >>>>>>>>>>>> attention to range switching. >>>>>>>>>>>> >>>>>>>>>>>> Not saying it can't be done, just that it's not nearly as simple as it >>>>>>>>>>>> sounds. >>>>>>>>>>> >>>>>>>>>>> I don't need amps--I'd be quite happy with a 50-100 mA top end. Most of >>>>>>>>>>> what I need a curve tracer for is measuring weird RF devices with lousy >>>>>>>>>>> datasheets. >>>>>>>>>>> >>>>>>>>>>> Cheers >>>>>>>>>>> >>>>>>>>>>> Phil Hobbs >>>>>>>>>> >>>>>>>>>> Oh foo, the fixturing for RF devices can get really weird and expensive. >>>>>>>>>> Those will be on you if we get together and create a USB curve tracer. >>>>>>>>>> >>>>>>>>>> ?-) >>>>>>>>> >>>>>>>>> The problem with a lot of (most) RF devices is their terrible DC >>>>>>>>> characterization. Like, no DC curves at all. The s-params give you a rough idea >>>>>>>>> of the high-speed behavior. >>>>>>>>> >>>>>>>>> RF data sheets often say to just adjust the bias until it works. The specified >>>>>>>>> breakdown voltage specs are not useful, because sometimes they assume, without >>>>>>>>> saying so, that the drain voltage can swing to multiples of the supply voltage, >>>>>>>>> and factor that in. Parts spec'd at 6 volts can cheerfully avalanche at 25. A >>>>>>>>> detector diode might be spec'd a 2 volts reverse, woth no curves at all, because >>>>>>>>> they figure that's all a detector user needs to know. >>>>>>>> >>>>>>>> I'm normally working down on the low frequency end anyway, e.g. a 100 >>>>>>>> MHz biochip front end using 14-GHz pHEMTs and 40-GHz bipolars, so simple >>>>>>>> concepts like C_JC and R_BB' still work OK. Boontons rule! >>>>>>> >>>>>>> Right. You have to make your own curves, like C-vs-V. That's why a USB >>>>>>> curve tracer should measure capacitances, too. >>>>>>> >>>>>>> Heck, somebody here with a curve tracer and a Booonton could set up a >>>>>>> garage operation to characterize parts for a fee. Modelithics does >>>>>>> that for big bucks. >>>>>>> >>>>>>> >>>>>> >>>>>> Have you measured the leakage of the Boonton's biasing system? Maybe >>>>>> one could just hang a little board on the back of one, and do it that way. >>>>>> >>>>>> Cheers >>>>>> >>>>>> Phil Hobbs >>>>> >>>>> Good idea. Design the USB curve tracer to run through a Boonton 72 to >>>>> get the capacitances. The Boonton has an analog output on the back. >>>>> But it would be more fun to do it ourselves. >>>>> >>>>> >>>> >>>> Sure, I'm in. Do you think it might actually sell? >>>> >>> >>> Well, some for sure. Ten a year? 200? >>> >>> Time to start scribbling. >> >> We might make two or more models. Some dependency on price, features, and >> performance for sales outside this ng. Within the ng we could probably >> make and deploy one to two dozen. Then with real world experience i >> think we could do a better job of selling it outside. >> >> ?-) > > How practical do you think it would it be to add temperature > characterization? I'd be interested in that.Just a SMOP. Measure Delta V_BE at three known currents, dancing back and forth as needed to take out the effects of drift and R_E. Should be able to get Theta_JA and maybe the thermal mass of the die + paddle. I'd like to be able to measure noise as well. It will be interesting to figure out how to make a test fixture that prevents all transistors from oscillating under all conditions. Maybe the thing to do is to sell the tracer cheap and make money off the disposable test boards. (Shades of early IBM.) 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 USA +1 845 480 2058 hobbs at electrooptical dot net http://electrooptical.net
Reply by ●February 16, 20132013-02-16
On Wed, 13 Feb 2013 22:10:49 -0800, Jon Kirwan <jonk@infinitefactors.org> wrote:> >> >>Ebbers-Moll is ok for bipolar but i would like to get Gummel-Poon =models.> >I agree. I was just pointing out that it starts out >interesting at the DC level and gets even more interesting >later. This is an area I really enjoy. > >Take VA. The easiest way is to extrapolate Ic vs Vce in the >CE configuration. But the error can be very high. Another way >is to measure the slope very near Vbc=3D0. Another way, though, >is using ln(Ic) vs Vbe at two different Vbc values, computing >it as a ratio. And it's better done that way. But that >assumes basewidth modulation is negligible. So additional >tests using different Vbe to see how well the assumption >holds may be useful. > >I'll be picking up a few BJT testers; 575, 577, and fairchild >6200, in a couple of months. > >>MOS models are a different subject that i didn't get to study. > >I'm in a similar boat. I've only basic knowledge here. They >don't seem as intriguing, though. > >JonWhile i may be ignorant of the MOS models there are many more of them, at least 9 IIRC. JT would know this well. As i understand it MOS 1 thru 3 are roughly of the grade of E-M 1 and 3 thru 6 span E-M 2 to G-P and 7 above account for even more device physics. ?-)
