Let's Take A Vote...
in [Design]
|
From: Grant on 7 Aug 2010 17:00 On Sat, 07 Aug 2010 10:18:17 -0700, John Larkin <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote: >On Sat, 7 Aug 2010 10:00:38 -0700 (PDT), j <jdc1789(a)gmail.com> wrote: > >>On Aug 7, 8:25 am, John Larkin >><jjlar...(a)highNOTlandTHIStechnologyPART.com> wrote: >>> On Sat, 07 Aug 2010 15:56:26 +0100, John Devereux >>> >>> >>> >>> >>> >>> <j...(a)devereux.me.uk> wrote: >>> >"markp" <map.nos...(a)f2s.com> writes: >>> >>> >> "John Devereux" <j...(a)devereux.me.uk> wrote in message >>> >>news:87mxth70en.fsf(a)devereux.me.uk... >>> >>> Jim Thompson <To-Email-Use-The-Envelope-I...(a)On-My-Web-Site.com> writes: >>> >>> >>>> Let's Take A Vote... >>> >>> >>>> While I write this up, hopefully sometime this weekend, let me ask for >>> >>>> votes... >>> >>> >>>> How many think, as Larkin opines, "charge is not conserved" ?? >>> >>> >>>> How many think charge IS conserved ?? >>> >>> >>>> Just curious what I'm up against here. >>> >>> >>> It depends on the context and your definitions, as already pointed out >>> >>> and which you still refuse to provide. Are you using a definition which >>> >>> says capacitors store charge or not? Is the quantity Q=CV to be regarded >>> >>> as "charge" or not? Is charge "delivered" or does it "flow through"? >>> >>> >>> *Without* any context, I would have said "charge is conserved". You >>> >>> don't need to spend three weeks proving this, just point to Kirchoff. >>> >>> >>> But the context of the original thread was all abouit switched >>> >>> capacitors and whether the "capacitors charge" was always conserved when >>> >>> transfered to another. We routinely refer to the quantity Q=CV as the >>> >>> capacitors "charge", it is this quantity which is not conserved, I.e., >>> >>> you can sum them before and after the switching operation and it is >>> >>> different. Not sophistry, just a normal use of terms in a circuit >>> >>> description. >>> >>> >>> And it is obvious that this was the intended usage, since otherwise the >>> >>> "charge of the capacitor" is always zero! >>> >>> >>> Basically, our routine use of the word is ambiguous, you can easily >>> >>> "prove someone wrong" by assuming the opposite usage to that intended, >>> >>> >>> -- >>> >>> >>> John Devereux >>> >>> >> Absolutely right. The way we talk about a 'capacitor's charge', using the >>> >> Q=CV equation, relates to the absolute value of charge on each plate, one >>> >> plate has +Q charge and the other -Q charge. The net charge on a capacitor >>> >> is zero (it has to be, since the same current goes into a capacitor as comes >>> >> out while 'charging' it, there can be no net gain or loss of charge inside >>> >> the capacitor). >>> >>> >> Therefore the 'capacitor's charge', by the definition above, is not >>> >> conserved, but the net charge is. When talking about charge conservation we >>> >> have to be careful about what our definitions of what we mean by 'charge' >>> >> are. >>> >>> >> I think John actually understands this, it's just the way it's put across >>> >> leads others to come to different conclusions. In a quote from a post from >>> >> the Magic Capacitors! thread he said to me: "We say that a capacitor stores >>> >> charge, the amount being C*V in coulombs, and it works. My whole point, >>> >> which has evoked such ranting, is that when you use this convention, be >>> >> careful about designing using the concept that (this kind of) charge is >>> >> always conserved." >>> >>> >Well Thank You and Kudos, Mark, for being one of the few people capable >>> >of changing their mind and admitting it. (About their interpretation of >>> >what someone else meant, not the physics itself!). >>> >>> "Charge " is just a word, and words are subject to different >>> definitions by different people. I don't think many people here would >>> get the reality wrong. >>> >>> Engineers don't always use terms or methods that scientists would >>> approve of. We do what works. When we do cheat the physics, we need to >>> be careful. >>> >>> John- Hide quoted text - >>> >>> - Show quoted text - >> >> >> >>"When we do cheat the physics" ... interesting concept. Well, I would have said, appear to cheat... >> > >Yup. We use math that doesn't actually work, but gets close enough to >do the job. We use terminology, like "charge", in our own way. We use >empirical properties of things for which we have no theory or >explanation. We do simulation and firmware where any concept of proper >dimensional analysis is totally lost... just sling the numbers around >until it works. We use proper physics when it helps, ignore it when it >doesn't. Or is irrelevant, same thing? > >What's interesting is that most physicists are terrible electronic >circuit designers, as their journals often show in hilarious ways. Like the circuit in Wireless World I saw when a teenager for biasing one of those new LEDs with a temperature compensated voltage supply, while the rest of the world simply put in a dropping resistor? I lost a bit of faith in published magazine circuits back then. > >One example is their devoted love of jfet differential pairs when the >second fet isn't necessary and just multiplies the noise by sqrt(2). Dunno that one, I didn't use jfets in my designs, closest was the ujt (? 2n6027) which is more like an anode gate scr. >And their abiding affection for trimpots, as opposed to proper biasing >design. Well, I put four in recently, two I could do in software (zero), but the other two (span) work for me 'cos it's an integer only PIC chip doing the numbers, and I can't rely on the uC to "spackle over the holes" ;^) I tend toward integer only fixed point numbers on 8bit controllers. And, I may have a processing issue measuring very close to zero that the zero trimpots will help me put a number on. Prototypes are like that, anticipate issues then find different stuff one didn't think of. Grant. > >John >
From: John Larkin on 7 Aug 2010 17:12
On Sun, 08 Aug 2010 07:00:45 +1000, Grant <omg(a)grrr.id.au> wrote: >On Sat, 07 Aug 2010 10:18:17 -0700, John Larkin <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote: > >>On Sat, 7 Aug 2010 10:00:38 -0700 (PDT), j <jdc1789(a)gmail.com> wrote: >> >>>On Aug 7, 8:25�am, John Larkin >>><jjlar...(a)highNOTlandTHIStechnologyPART.com> wrote: >>>> On Sat, 07 Aug 2010 15:56:26 +0100, John Devereux >>>> >>>> >>>> >>>> >>>> >>>> <j...(a)devereux.me.uk> wrote: >>>> >"markp" <map.nos...(a)f2s.com> writes: >>>> >>>> >> "John Devereux" <j...(a)devereux.me.uk> wrote in message >>>> >>news:87mxth70en.fsf(a)devereux.me.uk... >>>> >>> Jim Thompson <To-Email-Use-The-Envelope-I...(a)On-My-Web-Site.com> writes: >>>> >>>> >>>> Let's Take A Vote... >>>> >>>> >>>> While I write this up, hopefully sometime this weekend, let me ask for >>>> >>>> votes... >>>> >>>> >>>> How many think, as Larkin opines, "charge is not conserved" ?? >>>> >>>> >>>> How many think charge IS conserved ?? >>>> >>>> >>>> Just curious what I'm up against here. >>>> >>>> >>> It depends on the context and your definitions, as already pointed out >>>> >>> and which you still refuse to provide. Are you using a definition which >>>> >>> says capacitors store charge or not? Is the quantity Q=CV to be regarded >>>> >>> as "charge" or not? Is charge "delivered" or does it "flow through"? >>>> >>>> >>> *Without* any context, I would have said "charge is conserved". You >>>> >>> don't need to spend three weeks proving this, just point to Kirchoff. >>>> >>>> >>> But the context of the original thread was all abouit switched >>>> >>> capacitors and whether the "capacitors charge" was always conserved when >>>> >>> transfered to another. We routinely refer to the quantity Q=CV as the >>>> >>> capacitors "charge", it is this quantity which is not conserved, I.e., >>>> >>> you can sum them before and after the switching operation and it is >>>> >>> different. Not sophistry, just a normal use of terms in a circuit >>>> >>> description. >>>> >>>> >>> And it is obvious that this was the intended usage, since otherwise the >>>> >>> "charge of the capacitor" is always zero! >>>> >>>> >>> Basically, our routine use of the word is ambiguous, you can easily >>>> >>> "prove someone wrong" by assuming the opposite usage to that intended, >>>> >>>> >>> -- >>>> >>>> >>> John Devereux >>>> >>>> >> Absolutely right. The way we talk about a 'capacitor's charge', using the >>>> >> Q=CV equation, relates to the absolute value of charge on each plate, one >>>> >> plate has +Q charge and the other -Q charge. The net charge on a capacitor >>>> >> is zero (it has to be, since the same current goes into a capacitor as comes >>>> >> out while 'charging' it, there can be no net gain or loss of charge inside >>>> >> the capacitor). >>>> >>>> >> Therefore the 'capacitor's charge', by the definition above, is not >>>> >> conserved, but the net charge is. When talking about charge conservation we >>>> >> have to be careful about what our definitions of what we mean by 'charge' >>>> >> are. >>>> >>>> >> I think John actually understands this, it's just the way it's put across >>>> >> leads others to come to different conclusions. In a quote from a post from >>>> >> the Magic Capacitors! thread he said to me: "We say that a capacitor stores >>>> >> charge, the amount being C*V in coulombs, and it works. My whole point, >>>> >> which has evoked such ranting, is that when you use this convention, be >>>> >> careful about designing using the concept that (this kind of) charge is >>>> >> always conserved." >>>> >>>> >Well Thank You and Kudos, Mark, for being one of the few people capable >>>> >of changing their mind and admitting it. (About their interpretation of >>>> >what someone else meant, not the physics itself!). >>>> >>>> "Charge " is just a word, and words are subject to different >>>> definitions by different people. I don't think many people here would >>>> get the reality wrong. >>>> >>>> Engineers don't always use terms or methods that scientists would >>>> approve of. We do what works. When we do cheat the physics, we need to >>>> be careful. >>>> >>>> John- Hide quoted text - >>>> >>>> - Show quoted text - >>> >>> >>> >>>"When we do cheat the physics" ... interesting concept. > >Well, I would have said, appear to cheat... >>> >> >>Yup. We use math that doesn't actually work, but gets close enough to >>do the job. We use terminology, like "charge", in our own way. We use >>empirical properties of things for which we have no theory or >>explanation. We do simulation and firmware where any concept of proper >>dimensional analysis is totally lost... just sling the numbers around >>until it works. We use proper physics when it helps, ignore it when it >>doesn't. > >Or is irrelevant, same thing? >> >>What's interesting is that most physicists are terrible electronic >>circuit designers, as their journals often show in hilarious ways. > >Like the circuit in Wireless World I saw when a teenager for biasing >one of those new LEDs with a temperature compensated voltage supply, >while the rest of the world simply put in a dropping resistor? > >I lost a bit of faith in published magazine circuits back then. >> >>One example is their devoted love of jfet differential pairs when the >>second fet isn't necessary and just multiplies the noise by sqrt(2). > >Dunno that one, I didn't use jfets in my designs, closest was the ujt >(? 2n6027) which is more like an anode gate scr. > >>And their abiding affection for trimpots, as opposed to proper biasing >>design. > >Well, I put four in recently, two I could do in software (zero), but the >other two (span) work for me 'cos it's an integer only PIC chip doing >the numbers, and I can't rely on the uC to "spackle over the holes" ;^) > >I tend toward integer only fixed point numbers on 8bit controllers. I do almost my embedded programming in bare-metal assembly, using scaled integer math. Calibrations are usually fractional multiplies. On a recent product, one advertised voltage range was +-2.56. The actual ADC span was about +-2.7. We read the ADC, subtract an offset cal factor, then multiply by a gain cal factor. The gain cal is nominally 1.05. So we multiply by 0.05 and add that to the code. The 0.05 is expressed as a 16-bit fractional, and the multiply is just a 16x16==>32 followed by a 16-bit right shift. "0.05" is expressed as integer 3277. Of course, as you mention, codes are lost. One spacking technique is to oversample, calibrate, and lowpass filter. John |