Forums

SPICE gets it wrong

Started by Tim Williams May 7, 2017
On 05/10/2017 04:49 PM, Jim Thompson wrote:

>> BJTs are really more like voltage controlled current sources with a >> pathology than current controlled current sources > > Bwahahahahahaha >:-} > > ...Jim Thompson
There have been analog designers with a lot more experience than I have that've said much the same thing. The Ebers-Moll equation relating Ie to Vbe says nothing about Ie dependence on base current. Gummel-Poon charge-controlled model is useless outside of computer simulation
"bitrex" <bitrex@de.lete.earthlink.net> wrote in message 
news:YyKQA.72159$De7.40392@fx24.iad...
> BJTs are really more like voltage controlled current sources with a > pathology than current controlled current sources >
And with much more gain than MOSFETs! I like to think of low-Vce(sat) transistors as "MOSFETs with 5x more gain but a really leaky gate". They have a reasonably linear (including 3rd quadrant) "Rce(on)", too. "Vbe(th)" is also way more repeatable. I was testing almost a hundred boards, recently, and measured a current-limited output on each of them as part of the test spec. The design is guard-banded, so the current limit is high enough never to matter in normal operation, while it's also low enough to dissipate a safe amount of power when shorted. Despite the wide design tolerance, they all measured within 0.3% of each other. (Evidently, room temperature was pretty stable between those days of testing.) Tim -- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com
On a sunny day (Wed, 10 May 2017 16:16:56 -0400) it happened bitrex
<bitrex@de.lete.earthlink.net> wrote in <YyKQA.72159$De7.40392@fx24.iad>:

>BJTs are really more like voltage controlled current sources with a >pathology than current controlled current sources
Well, maybe you misread the label, and it was a FET... :-)
On a sunny day (Wed, 10 May 2017 11:30:14 -0400) it happened bitrex
<bitrex@de.lete.earthlink.net> wrote in <amGQA.20524$rp2.9505@fx31.iad>:

>On 05/10/2017 10:42 AM, Jan Panteltje wrote: > >>>> I exected a circuit diagram ;-( >>> >>> Sorry, that's too valuable to give away. I will say: drive the gates >>> hard and ignore the data sheets. >> >> Let me guess, you used an ethernet transformer...? > >Nah nah it was an _isdn_ xfrmr. That's the good shit > >> >> Maybe some science field, was just reading CERN started a new linear accelerator: >> https://home.cern/about/updates/2017/05/brand-new-linear-accelerator-cern >> > >Where's Famous Chef Pante?
Na, was in the kitchen just now baking mushrooms... MY opinion on this, and actually last night there was, on cable here, a professor giving college about CERN comparing it to a super microscope. You know ever higher energies needed to take a picture of ever smaller things. But if you ask me, I did say: 'Do you really expect to get a deeper insight into the Tesla autopilot software by driving those into each other at 1000 km/h?' So..
On 05/10/2017 09:43 PM, Tim Williams wrote:
> "bitrex" <bitrex@de.lete.earthlink.net> wrote in message > news:YyKQA.72159$De7.40392@fx24.iad... >> BJTs are really more like voltage controlled current sources with a >> pathology than current controlled current sources >> > > And with much more gain than MOSFETs! > > I like to think of low-Vce(sat) transistors as "MOSFETs with 5x more > gain but a really leaky gate". They have a reasonably linear (including > 3rd quadrant) "Rce(on)", too. > > "Vbe(th)" is also way more repeatable. I was testing almost a hundred > boards, recently, and measured a current-limited output on each of them > as part of the test spec. The design is guard-banded, so the current > limit is high enough never to matter in normal operation, while it's > also low enough to dissipate a safe amount of power when shorted. > Despite the wide design tolerance, they all measured within 0.3% of each > other. (Evidently, room temperature was pretty stable between those > days of testing.) > > Tim >
Ya, I suppose if you could design a transconductance physics box with an exponential law instead of square that also had a pretty well-defined "gate" threshold of 0.6 volts and drew no "gate" current, there would be little reason to use anything else assuming similar noise performance over a similar bandwidth. I think (x)HEMTs come pretty close, sadly they are not literally a dime a dozen and likely never will be.