From: Jim Thompson on 4 Jul 2010 18:11 On Sat, 3 Jul 2010 22:07:56 -0700 (PDT), dagmargoodboat(a)yahoo.com wrote: >I need a current limiter, so I looked at poly fuses. They're pretty >sad. And regular fuses are awfully crude--make one mistake and you >have to desolder a fuse? That's no fun. > >So I was thinking about the Larksonian current limiter... > >Fig. 1 (View in Courier font) >====== > Q1 > >-------. .-------> > V / > ------- > | > | > | > .-. > | |R1 > | | > '-' > | > | > === > >...which was offered here as kind of a half-serious lark by a certain >John Larkin. > >As John pointed out, Q1 can be a 2:1 Hfe-binned part, giving a 2:1 >limit tolerance, which isn't too bad for a non-critical application. > > i.limit ~= Hfe * i.b > >So, I'm supplying a little +5V to the real world via a small >connector, to communicate with and power a small external device. I'm >not sure what the draw is, maybe 50-100mA, max., but I do want to >supply the full +5V and not a lot less--the voltage drop should be >low. > >John's circuit's tolerances are fine--I'd set it to 200mA or so, and >get 150-300mA. That's great. But if the output's shorted, that's >possibly 5v x 300mA--too much dissipation for Q1. > > >So I added foldback-- > >Fig. 2 >====== > Q2 > >--+-----. .-------+----> > | V / | > | -------- | > | | | > '>|Q3 | R2 | > |------|----/\/\/--' > /| | > | | > '--------+ > | > .-. > | |R3 > | | > '-' > | > | > === > >There, that's better. If Q2 ever desats, Q3 comes on, robs Q2's base >drive, and the output collapses. Once shorted, an idling current >flows through Q3 e-b and R2 to the load, so the thing restarts once >the short is removed. > >(The classic foldback uses a divider from Q2(c) to GND, tap goes to >Q3(b). I didn't do that here--we don't need that much of a hair- >trigger.) > >Hmmmm. > >It's pretty fast. Maybe a cap to slow down Q3, so we can tolerate a >brief spike without triggering... > > >Fig. 3 >====== > Q4 > >--+-----+---. .--------+----> > | | V / | > | C1 --- -------- | > | --- | | > '>| | | R2 | > Q5 |---+-------|---/\/\/---' > /| | > | | > '-------------+ > | > .-. > | |R3 > | | > '-' > | > === > >Hmmmmmm. > > >--James Arthur Or do it right... http://analog-innovations.com/SED/CurrentSense.pdf ...Jim Thompson -- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | Obama: A reincarnation of Nixon, narcissistically posing in politically-correct black-face, but with fewer scruples.
From: dagmargoodboat on 4 Jul 2010 18:28 On Jul 4, 2:44 pm, John Larkin <jjlar...(a)highNOTlandTHIStechnologyPART.com> wrote: > On Sun, 4 Jul 2010 10:10:57 -0700 (PDT), dagmargoodb...(a)yahoo.com > wrote: <snip> > >Meanwhile I was going exactly the opposite way, trying to minimize the > >transistor dependencies. I was just looking at some BJT datasheets. > >The main problem is d(Hfe)/dt. > > >Here's a compromise: > > >Fig. 6 > >====== > > Q60 R60 > > >--+------+---. .--/\/\/---+---> > > | | V / | > > | C60 --- -------- | > > | --- | | > > '>| | | R61 | > > Q61 |----+-------|----/\/\/----' > > /| | > > | | > > '--------------+ > > | > > .-. > > | | R62 > > | | > > '-' > > | > > === > > >It's simple, self-resetting, drifty and crude, but R60 tames the > >transistor dependencies considerably. > > >To keep droop reasonable I'd set the limit waayyy high, so the 0.6v > >trigger point doesn't happen until the load is already majorly > >pathological. > > If you're willing to put a sense resistor in the current path, well, > you're cheating. You may as well do a vanilla 2-transistor foldback > current limiter. Adding a schottky diode can reduce the voltage loss > to a few tenths. > > John By "vanilla" do you mean an emitter-follower with a sense transistor that robs its base-drive? James
From: dagmargoodboat on 4 Jul 2010 19:05 On Jul 4, 5:11 pm, Jim Thompson <To-Email-Use-The-Envelope-I...(a)On-My- Web-Site.com> wrote: > On Sat, 3 Jul 2010 22:07:56 -0700 (PDT), dagmargoodb...(a)yahoo.com > wrote: > > > > >I need a current limiter, so I looked at poly fuses. They're pretty > >sad. And regular fuses are awfully crude--make one mistake and you > >have to desolder a fuse? That's no fun. > > >So I was thinking about the Larksonian current limiter... > > >Fig. 1 (View in Courier font) > >====== > > Q1 > > >-------. .-------> > > V / > > ------- > > | > > | > > | > > .-. > > | |R1 > > | | > > '-' > > | > > | > > === > > >...which was offered here as kind of a half-serious lark by a certain > >John Larkin. > > >As John pointed out, Q1 can be a 2:1 Hfe-binned part, giving a 2:1 > >limit tolerance, which isn't too bad for a non-critical application. > > > i.limit ~= Hfe * i.b > > >So, I'm supplying a little +5V to the real world via a small > >connector, to communicate with and power a small external device. I'm > >not sure what the draw is, maybe 50-100mA, max., but I do want to > >supply the full +5V and not a lot less--the voltage drop should be > >low. > > >John's circuit's tolerances are fine--I'd set it to 200mA or so, and > >get 150-300mA. That's great. But if the output's shorted, that's > >possibly 5v x 300mA--too much dissipation for Q1. > > >So I added foldback-- > > >Fig. 2 > >====== > > Q2 > > >--+-----. .-------+----> > > | V / | > > | -------- | > > | | | > > '>|Q3 | R2 | > > |------|----/\/\/--' > > /| | > > | | > > '--------+ > > | > > .-. > > | |R3 > > | | > > '-' > > | > > | > > === > > >There, that's better. If Q2 ever desats, Q3 comes on, robs Q2's base > >drive, and the output collapses. Once shorted, an idling current > >flows through Q3 e-b and R2 to the load, so the thing restarts once > >the short is removed. > > >(The classic foldback uses a divider from Q2(c) to GND, tap goes to > >Q3(b). I didn't do that here--we don't need that much of a hair- > >trigger.) > > >Hmmmm. > > >It's pretty fast. Maybe a cap to slow down Q3, so we can tolerate a > >brief spike without triggering... > > >Fig. 3 > >====== > > Q4 > > >--+-----+---. .--------+----> > > | | V / | > > | C1 --- -------- | > > | --- | | > > '>| | | R2 | > > Q5 |---+-------|---/\/\/---' > > /| | > > | | > > '-------------+ > > | > > .-. > > | |R3 > > | | > > '-' > > | > > === > > >Hmmmmmm. > > >--James Arthur > > Or do it right... > > http://analog-innovations.com/SED/CurrentSense.pdf That logic-output current-sensor's pretty close technique-wise to my electronic circuit-breaker of Fig. 5 above, except for the method of generating the less-than-Vbe offset voltage for the threshold. That's cute. I recognize it as an old analog IC trick, but I've never done it. I like the hysteresis too. Thanks. -- Cheers, James Arthur
From: John Larkin on 4 Jul 2010 20:14 On Sun, 4 Jul 2010 15:28:33 -0700 (PDT), dagmargoodboat(a)yahoo.com wrote: >On Jul 4, 2:44�pm, John Larkin ><jjlar...(a)highNOTlandTHIStechnologyPART.com> wrote: >> On Sun, 4 Jul 2010 10:10:57 -0700 (PDT), dagmargoodb...(a)yahoo.com >> wrote: > ><snip> > >> >Meanwhile I was going exactly the opposite way, trying to minimize the >> >transistor dependencies. �I was just looking at some BJT datasheets. >> >The main problem is d(Hfe)/dt. >> >> >Here's a compromise: >> >> >Fig. 6 >> >====== >> > � � � � � � � � �Q60 � � �R60 >> > �>--+------+---. � � �.--/\/\/---+---> >> > � � | � � �| � �V � �/ � � � � � | >> > � � | C60 --- �-------- � � � � �| >> > � � | � � --- � � �| � � � � � � | >> > � � '>| � �| � � � | � � R61 � � | >> > �Q61 �|----+-------|----/\/\/----' >> > � � �/| � � � � � �| >> > � � | � � � � � � �| >> > � � '--------------+ >> > � � � � � � � � � �| >> > � � � � � � � � � .-. >> > � � � � � � � � � | | R62 >> > � � � � � � � � � | | >> > � � � � � � � � � '-' >> > � � � � � � � � � �| >> > � � � � � � � � � === >> >> >It's simple, self-resetting, drifty and crude, but R60 tames the >> >transistor dependencies considerably. >> >> >To keep droop reasonable I'd set the limit waayyy high, so the 0.6v >> >trigger point doesn't happen until the load is already majorly >> >pathological. >> >> If you're willing to put a sense resistor in the current path, well, >> you're cheating. You may as well do a vanilla 2-transistor foldback >> current limiter. Adding a schottky diode can reduce the voltage loss >> to a few tenths. >> >> John > >By "vanilla" do you mean an emitter-follower with a sense transistor >that robs its base-drive? > >James Right. Add a schottky to take away some of the Vbe requirement, and a couple more resistors to add foldback. Or go with an opamp and a small-value sense resistor, and a p-channel fet. Or maybe one of these... ftp://jjlarkin.lmi.net/Goofy_Ilims.JPG The polyfuse is looking better and better. Foldbacks can have strange behavior with some non-resistive loads. That's why a shutdown with periodic retry is good. Hey, have a uP port pin turn on a series PNP or pfet, and use a spare ADC channel to digitize the load voltage. Then do the math. I'm doing some fairly complex shutdowns in software on the 4-20 mA gadget we're doing... software lowpass filtering, trip thresholds, automatic retry, stuff like that. Another thing I like to do is measure power fet voltage and current and heatsink temp, run a realtime simulation of thermal dynamics, and shut down based on junction temperature. John
From: John Larkin on 4 Jul 2010 20:22
On Sun, 4 Jul 2010 16:05:10 -0700 (PDT), dagmargoodboat(a)yahoo.com wrote: >On Jul 4, 5:11�pm, Jim Thompson <To-Email-Use-The-Envelope-I...(a)On-My- >Web-Site.com> wrote: >> On Sat, 3 Jul 2010 22:07:56 -0700 (PDT), dagmargoodb...(a)yahoo.com >> wrote: >> >> >> >> >I need a current limiter, so I looked at poly fuses. �They're pretty >> >sad. �And regular fuses are awfully crude--make one mistake and you >> >have to desolder a fuse? �That's no fun. >> >> >So I was thinking about the Larksonian current limiter... >> >> >Fig. 1 � � (View in Courier font) >> >====== >> > � � � � � � Q1 >> > �>-------. � � .-------> >> > � � � � � V � / >> > � � � � �------- >> > � � � � � � | >> > � � � � � � | >> > � � � � � � | >> > � � � � � �.-. >> > � � � � � �| |R1 >> > � � � � � �| | >> > � � � � � �'-' >> > � � � � � � | >> > � � � � � � | >> > � � � � � �=== >> >> >...which was offered here as kind of a half-serious lark by a certain >> >John Larkin. >> >> >As John pointed out, Q1 can be a 2:1 Hfe-binned part, giving a 2:1 >> >limit tolerance, which isn't too bad for a non-critical application. >> >> > �i.limit ~= Hfe * i.b >> >> >So, I'm supplying a little +5V to the real world via a small >> >connector, to communicate with and power a small external device. �I'm >> >not sure what the draw is, maybe 50-100mA, max., but I do want to >> >supply the full +5V and not a lot less--the voltage drop should be >> >low. >> >> >John's circuit's tolerances are fine--I'd set it to 200mA or so, and >> >get 150-300mA. �That's great. �But if the output's shorted, that's >> >possibly 5v x 300mA--too much dissipation for Q1. >> >> >So I added foldback-- >> >> >Fig. 2 >> >====== >> > � � � � � � �Q2 >> > �>--+-----. � � �.-------+----> >> > � � | � � �V � �/ � � � �| >> > � � | � � -------- � � � | >> > � � | � � � �| � � � � � | >> > � � '>|Q3 � �| � � �R2 � | >> > � � � |------|----/\/\/--' >> > � � �/| � � �| >> > � � | � � � �| >> > � � '--------+ >> > � � � � � � �| >> > � � � � � � .-. >> > � � � � � � | |R3 >> > � � � � � � | | >> > � � � � � � '-' >> > � � � � � � �| >> > � � � � � � �| >> > � � � � � � === >> >> >There, that's better. �If Q2 ever desats, Q3 comes on, robs Q2's base >> >drive, and the output collapses. �Once shorted, an idling current >> >flows through Q3 e-b and R2 to the load, so the thing restarts once >> >the short is removed. >> >> >(The classic foldback uses a divider from Q2(c) to GND, tap goes to >> >Q3(b). �I didn't do that here--we don't need that much of a hair- >> >trigger.) >> >> >Hmmmm. >> >> >It's pretty fast. �Maybe a cap to slow down Q3, so we can tolerate a >> >brief spike without triggering... >> >> >Fig. 3 >> >====== >> > � � � � � � � � �Q4 >> > �>--+-----+---. � � �.--------+----> >> > � � | � � | � �V � �/ � � � � | >> > � � | C1 --- �-------- � � � �| >> > � � | � �--- � � �| � � � � � | >> > � � '>| � | � � � | � � R2 � �| >> > � Q5 �|---+-------|---/\/\/---' >> > � � �/| � � � � � | >> > � � | � � � � � � | >> > � � '-------------+ >> > � � � � � � � � � | >> > � � � � � � � � �.-. >> > � � � � � � � � �| |R3 >> > � � � � � � � � �| | >> > � � � � � � � � �'-' >> > � � � � � � � � � | >> > � � � � � � � � �=== >> >> >Hmmmmmm. >> >> >--James Arthur >> >> Or do it right... >> >> http://analog-innovations.com/SED/CurrentSense.pdf > > >That logic-output current-sensor's pretty close technique-wise to my >electronic circuit-breaker of Fig. 5 above, except for the method of >generating the less-than-Vbe offset voltage for the threshold. That's >cute. I recognize it as an old analog IC trick, but I've never done >it. I like the hysteresis too. It just doesn't do the electronic fuse function. John |