feedback and stability
in [Design]
|
From: Michael Robinson on 22 Dec 2009 16:56 "ehsjr" <ehsjr(a)nospamverizon.net> wrote in message news:hgp586$r12$1(a)news.eternal-september.org... > Michael Robinson wrote: >> "Tim Wescott" <tim(a)seemywebsite.com> wrote in message >> news:K56dnfGL45LFBLPWnZ2dnUVZ_vRi4p2d(a)web-ster.com... >> >>>On Sun, 20 Dec 2009 15:32:04 -0500, Michael Robinson wrote: >>> >>> >>>>I built a voltage regulator for an old truck I own. The circuit uses a >>>>voltage controlled oscillator with a square wave output to drive the >>>>gate of a mosfet that controls the alternator's field. The voltage >>>>regulation feedback loop uses a voltage divider from the alternator >>>>output/vehicle battery, connected to the reference pin of a TL431. The >>>>TL431 controls the vco. >>>> >>>>Now, for a bit of a digression. Years ago I built a boost converter with >>>>a similar feedback loop, the major difference being that it used a bjt >>>>instead of a TL431. But it worked on the same principle, with a voltage >>>>divider from the output to the base of the bjt, which controlled a >>>>blocking oscillator. The output voltage wandered up and down and was >>>>very unstable, so I put a capacitor from the collector to the base of >>>>the bjt. That fixed the problem and made the output voltage stable. >>>> >>>>So, having had this experience, I put a cap from the cathode to the >>>>reference of the TL431 in my truck regulator. When I installed it in the >>>>truck, the alternator output regulation was unstable; the voltage jumped >>>>up and down like crazy. After much head scratching, on a hunch I took >>>>wire clippers and removed that cap from the circuit board. Then the >>>>regulator worked GREAT. System voltage perfectly stable, good load >>>>regulation. >>>> >>>>Perhaps I could be forgiven for thinking the boost converter and the >>>>vehicle voltage regulator would operate on the same principle and that a >>>>cap would have the same effect, in view of the fact that in both >>>>circuits the cap provides negative feedback in the controlling element >>>>of the feedback loop. But the two circuits have opposite behavior, and >>>>adding a cap to the controlling element in the feedback loop has exactly >>>>the opposite effect in each circuit. The boost regulator was unstable >>>>without a capacitor, and stable with one. Contrarily, the alternator >>>>regulator was stable without a capacitor, and unstable with one. >>>> >>>>At best I have some intuitive sense that the presence of a big >>>>mechanical device that drives the load and resides within the loop >>>>changes things dramatically. But I'm lacking in knowledge of theory and >>>>could use elucidation. Any pointers? >>> >>>A schematic would help - either an ASCII art one, or put one up on a >>>website somewhere so folks could look. >>> >>>Yes, the dynamics of your alternator are likely to be different from the >>>dynamics of your boost converter output stage, so that could certainly >>>have an effect on stability. Without knowing more about what you did >>>(and probably more about the alternator's dynamics) I can't say much. >>> >> >> >> I don't think I can find the diagram for the circuit, but the following >> should help: >> Here's a spice file for a vco with the topology of the one I used. I >> used a different comparator, and component values may be different, but I >> didn't change the basic design. >> This spice file was actually posted by John Popelish a few years ago, and >> I based my circuit on it. >> >> Version 4 >> SHEET 1 880 680 >> WIRE -16 -80 -128 -80 >> WIRE 304 -80 64 -80 >> WIRE -64 16 -80 16 >> WIRE 32 16 16 16 >> WIRE 304 48 304 -80 >> WIRE 336 48 336 16 >> WIRE 336 48 304 48 >> WIRE 336 80 336 48 >> WIRE 288 160 224 160 >> WIRE 224 176 224 160 >> WIRE -240 240 -304 240 >> WIRE -128 240 -128 -80 >> WIRE -128 240 -160 240 >> WIRE -16 240 -128 240 >> WIRE 96 256 96 16 >> WIRE 96 256 48 256 >> WIRE 128 256 96 256 >> WIRE 176 256 128 256 >> WIRE -80 272 -80 16 >> WIRE -48 272 -48 176 >> WIRE -48 272 -80 272 >> WIRE -16 272 -48 272 >> WIRE -48 304 -48 272 >> WIRE -128 320 -128 240 >> FLAG 432 288 0 >> FLAG 432 208 p12 >> FLAG 336 176 0 >> FLAG 224 272 0 >> FLAG 16 288 0 >> FLAG -48 384 0 >> FLAG 336 -64 p12 >> FLAG -48 96 p12 >> FLAG 16 224 p12 >> FLAG 128 176 p12 >> FLAG 224 80 p12 >> FLAG -304 320 0 >> FLAG -128 384 0 >> FLAG 336 48 Out >> SYMBOL voltage 432 192 R0 >> WINDOW 123 0 0 Left 0 >> WINDOW 39 0 0 Left 0 >> SYMATTR InstName V1 >> SYMATTR Value 12 >> SYMBOL Comparators\\LT1017 16 256 R0 >> SYMATTR InstName U1 >> SYMBOL nmos 176 176 R0 >> SYMATTR InstName M1 >> SYMATTR Value BSS145 >> SYMBOL nmos 288 80 R0 >> SYMATTR InstName M2 >> SYMATTR Value FDS6680A >> SYMBOL res 208 64 R0 >> SYMATTR InstName R1 >> SYMATTR Value 1k >> SYMBOL res 320 -80 R0 >> SYMATTR InstName R2 >> SYMATTR Value 1 >> SYMBOL res -64 288 R0 >> SYMATTR InstName R3 >> SYMATTR Value 100k >> SYMBOL res -64 80 R0 >> SYMATTR InstName R4 >> SYMATTR Value 100k >> SYMBOL res 112 160 R0 >> SYMATTR InstName R5 >> SYMATTR Value 10k >> SYMBOL cap -144 320 R0 >> SYMATTR InstName C1 >> SYMATTR Value 220n >> SYMBOL cap 96 0 R90 >> WINDOW 0 0 32 VBottom 0 >> WINDOW 3 32 32 VTop 0 >> SYMATTR InstName C3 >> SYMATTR Value 10n >> SYMBOL res 32 0 R90 >> WINDOW 0 0 56 VBottom 0 >> WINDOW 3 32 56 VTop 0 >> SYMATTR InstName R6 >> SYMATTR Value 100k >> SYMBOL res 80 -96 R90 >> WINDOW 0 0 56 VBottom 0 >> WINDOW 3 32 56 VTop 0 >> SYMATTR InstName R7 >> SYMATTR Value 100k >> SYMBOL res -144 224 R90 >> WINDOW 0 0 56 VBottom 0 >> WINDOW 3 32 56 VTop 0 >> SYMATTR InstName R8 >> SYMATTR Value 100k >> SYMBOL voltage -304 224 R0 >> WINDOW 123 0 0 Left 0 >> WINDOW 39 0 0 Left 0 >> SYMATTR InstName V2 >> SYMATTR Value PULSE(0 12 0 1) >> TEXT 184 360 Left 0 !.tran 1 >> >> At the input to the vco there's a TL431 with a pullup resistor on the >> cathode, and the cathode is connected to the vco input. >> For what it's worth, the truck is a 1994 Dodge 2500 Ram diesel. >> >> >> Putting a cap across the TL431 cathode/reference terminals ought to slow >> the circuit's response, which could be a problem. >> From the way the voltage was jumping around, it was something more than >> just loose regulation -- it's like the cap introduced some kind of >> nonlinear effect in the feedback loop's dynamics. It wasn't just >> drifting, it was bouncing around. >> >> > > > Drawing it out will make it clear. The first schematic below > is a "sanity check", just to make sure that the circuit > you described is something like it: > > + ----+----[PASS]------------+-----------------> field > | | | > +-----[VCO]---gnd | > | | | > +--[R]--+------+ | > | | | > __|__/ [C] [R] > / \ | | > /___\----+-------+ > | | > | [R] > | | > Gnd ----------+--------------+ > > Assuming the above captures the essence of what you have, read on. No, it looks more like this: field | mosfet | vco batt | | +12v---[R]---+------+ | | | | __|__/ [C] [R] / / \ | | /___\----+-------+ | | | [R] | | Gnd ----------+--------------+
From: Michael Robinson on 22 Dec 2009 17:23 "ehsjr" <ehsjr(a)nospamverizon.net> wrote in message news:hgp586$r12$1(a)news.eternal-september.org... > Michael Robinson wrote: >> "Tim Wescott" <tim(a)seemywebsite.com> wrote in message >> news:K56dnfGL45LFBLPWnZ2dnUVZ_vRi4p2d(a)web-ster.com... >> >>>On Sun, 20 Dec 2009 15:32:04 -0500, Michael Robinson wrote: >>> >>> >>>>I built a voltage regulator for an old truck I own. The circuit uses a >>>>voltage controlled oscillator with a square wave output to drive the >>>>gate of a mosfet that controls the alternator's field. The voltage >>>>regulation feedback loop uses a voltage divider from the alternator >>>>output/vehicle battery, connected to the reference pin of a TL431. The >>>>TL431 controls the vco. >>>> >>>>Now, for a bit of a digression. Years ago I built a boost converter with >>>>a similar feedback loop, the major difference being that it used a bjt >>>>instead of a TL431. But it worked on the same principle, with a voltage >>>>divider from the output to the base of the bjt, which controlled a >>>>blocking oscillator. The output voltage wandered up and down and was >>>>very unstable, so I put a capacitor from the collector to the base of >>>>the bjt. That fixed the problem and made the output voltage stable. >>>> >>>>So, having had this experience, I put a cap from the cathode to the >>>>reference of the TL431 in my truck regulator. When I installed it in the >>>>truck, the alternator output regulation was unstable; the voltage jumped >>>>up and down like crazy. After much head scratching, on a hunch I took >>>>wire clippers and removed that cap from the circuit board. Then the >>>>regulator worked GREAT. System voltage perfectly stable, good load >>>>regulation. >>>> >>>>Perhaps I could be forgiven for thinking the boost converter and the >>>>vehicle voltage regulator would operate on the same principle and that a >>>>cap would have the same effect, in view of the fact that in both >>>>circuits the cap provides negative feedback in the controlling element >>>>of the feedback loop. But the two circuits have opposite behavior, and >>>>adding a cap to the controlling element in the feedback loop has exactly >>>>the opposite effect in each circuit. The boost regulator was unstable >>>>without a capacitor, and stable with one. Contrarily, the alternator >>>>regulator was stable without a capacitor, and unstable with one. >>>> >>>>At best I have some intuitive sense that the presence of a big >>>>mechanical device that drives the load and resides within the loop >>>>changes things dramatically. But I'm lacking in knowledge of theory and >>>>could use elucidation. Any pointers? >>> >>>A schematic would help - either an ASCII art one, or put one up on a >>>website somewhere so folks could look. >>> >>>Yes, the dynamics of your alternator are likely to be different from the >>>dynamics of your boost converter output stage, so that could certainly >>>have an effect on stability. Without knowing more about what you did >>>(and probably more about the alternator's dynamics) I can't say much. >>> >> >> >> I don't think I can find the diagram for the circuit, but the following >> should help: >> Here's a spice file for a vco with the topology of the one I used. I >> used a different comparator, and component values may be different, but I >> didn't change the basic design. >> This spice file was actually posted by John Popelish a few years ago, and >> I based my circuit on it. >> >> Version 4 >> SHEET 1 880 680 >> WIRE -16 -80 -128 -80 >> WIRE 304 -80 64 -80 >> WIRE -64 16 -80 16 >> WIRE 32 16 16 16 >> WIRE 304 48 304 -80 >> WIRE 336 48 336 16 >> WIRE 336 48 304 48 >> WIRE 336 80 336 48 >> WIRE 288 160 224 160 >> WIRE 224 176 224 160 >> WIRE -240 240 -304 240 >> WIRE -128 240 -128 -80 >> WIRE -128 240 -160 240 >> WIRE -16 240 -128 240 >> WIRE 96 256 96 16 >> WIRE 96 256 48 256 >> WIRE 128 256 96 256 >> WIRE 176 256 128 256 >> WIRE -80 272 -80 16 >> WIRE -48 272 -48 176 >> WIRE -48 272 -80 272 >> WIRE -16 272 -48 272 >> WIRE -48 304 -48 272 >> WIRE -128 320 -128 240 >> FLAG 432 288 0 >> FLAG 432 208 p12 >> FLAG 336 176 0 >> FLAG 224 272 0 >> FLAG 16 288 0 >> FLAG -48 384 0 >> FLAG 336 -64 p12 >> FLAG -48 96 p12 >> FLAG 16 224 p12 >> FLAG 128 176 p12 >> FLAG 224 80 p12 >> FLAG -304 320 0 >> FLAG -128 384 0 >> FLAG 336 48 Out >> SYMBOL voltage 432 192 R0 >> WINDOW 123 0 0 Left 0 >> WINDOW 39 0 0 Left 0 >> SYMATTR InstName V1 >> SYMATTR Value 12 >> SYMBOL Comparators\\LT1017 16 256 R0 >> SYMATTR InstName U1 >> SYMBOL nmos 176 176 R0 >> SYMATTR InstName M1 >> SYMATTR Value BSS145 >> SYMBOL nmos 288 80 R0 >> SYMATTR InstName M2 >> SYMATTR Value FDS6680A >> SYMBOL res 208 64 R0 >> SYMATTR InstName R1 >> SYMATTR Value 1k >> SYMBOL res 320 -80 R0 >> SYMATTR InstName R2 >> SYMATTR Value 1 >> SYMBOL res -64 288 R0 >> SYMATTR InstName R3 >> SYMATTR Value 100k >> SYMBOL res -64 80 R0 >> SYMATTR InstName R4 >> SYMATTR Value 100k >> SYMBOL res 112 160 R0 >> SYMATTR InstName R5 >> SYMATTR Value 10k >> SYMBOL cap -144 320 R0 >> SYMATTR InstName C1 >> SYMATTR Value 220n >> SYMBOL cap 96 0 R90 >> WINDOW 0 0 32 VBottom 0 >> WINDOW 3 32 32 VTop 0 >> SYMATTR InstName C3 >> SYMATTR Value 10n >> SYMBOL res 32 0 R90 >> WINDOW 0 0 56 VBottom 0 >> WINDOW 3 32 56 VTop 0 >> SYMATTR InstName R6 >> SYMATTR Value 100k >> SYMBOL res 80 -96 R90 >> WINDOW 0 0 56 VBottom 0 >> WINDOW 3 32 56 VTop 0 >> SYMATTR InstName R7 >> SYMATTR Value 100k >> SYMBOL res -144 224 R90 >> WINDOW 0 0 56 VBottom 0 >> WINDOW 3 32 56 VTop 0 >> SYMATTR InstName R8 >> SYMATTR Value 100k >> SYMBOL voltage -304 224 R0 >> WINDOW 123 0 0 Left 0 >> WINDOW 39 0 0 Left 0 >> SYMATTR InstName V2 >> SYMATTR Value PULSE(0 12 0 1) >> TEXT 184 360 Left 0 !.tran 1 >> >> At the input to the vco there's a TL431 with a pullup resistor on the >> cathode, and the cathode is connected to the vco input. >> For what it's worth, the truck is a 1994 Dodge 2500 Ram diesel. >> >> >> Putting a cap across the TL431 cathode/reference terminals ought to slow >> the circuit's response, which could be a problem. >> From the way the voltage was jumping around, it was something more than >> just loose regulation -- it's like the cap introduced some kind of >> nonlinear effect in the feedback loop's dynamics. It wasn't just >> drifting, it was bouncing around. >> >> > > > Drawing it out will make it clear. The first schematic below > is a "sanity check", just to make sure that the circuit > you described is something like it: As I explained from the beginning, the votlage divider at the reference pin is connected to the battery/alternator output, like this: field | mosfet | vco batt | | +12v---[R]---+------+ | | | | __|__/ [C] [R] / / \ | | /___\----+-------+ | | | [R] | | Gnd ----------+--------------+ Can you see how it would regulate voltage? Any variation in the battery voltage will cause the cathode to pull the vco up or down, changing the duty cycle of the mosfet. It's a big feedback loop that includes the alternator. I want to know how the cap throws everything off.
From: Dave Platt on 22 Dec 2009 18:42 In article <AIbYm.104465$Wf2.40992(a)newsfe23.iad>, Michael Robinson <nospam(a)billburg.com> wrote: >As I explained from the beginning, the votlage divider at the reference pin >is connected to the battery/alternator output, like this: > > field > | > mosfet > | > vco batt > | | > +12v---[R]---+------+ | > | | | > __|__/ [C] [R] > / / \ | | > /___\----+-------+ > | | > | [R] > | | > Gnd ----------+--------------+ > >Can you see how it would regulate voltage? Any variation in the battery >voltage will cause the cathode to pull the vco up or down, changing the duty >cycle of the mosfet. It's a big feedback loop that includes the alternator. >I want to know how the cap throws everything off. It seems to me that the presence of the cap will necessarily delay the "visibility" of any change in the battery terminal voltage. The cap will tend to hold the ref/C/R/R junction at a constant voltage, even if the battery changes voltage rather abruptly and substantially. This means that the VCO's ability to react (increasing the MOSFET duty cycle and the field) will be seriously compromised. It'll be slow in reacting (and increasing the field) when the battery voltage drops. Similarly, it will be slow in reacting and decreasing the field, when the battery voltage climbs too high. I think you can get (and have gotten!) yourself into trouble, in any situation where the RC time constant created here, is slower than the speed at which the field strength (and alternator output voltage) reacts to changes in the VCO control voltage. The battery voltage will tend to overshoot in both directions, because the control circuit is *always* trying to catch up with (and react to) the overly-strong reaction to its previous change. At best, the output will ring. At worst, it will oscillate. This is not dissimilar to a classic stability problem in audio amplifiers... trying to use a fast front end circuit, with slow power-stage transistors. Using a fast-acting control system, with a lot of gain, in a slow feedback loop, is a common recipe for serious instability. I think that in order to ensure stability, you need to make sure that the RC time constant involved isn't larger than the amount of time required for the changes in VCO control voltage to take effect at the battery. Another approach might be to reduce the gain of the control system... i.e. the amount by which a change in the VCO control voltage affects the duty cycle of the alternator. Someone with more EE knowledge than I, might be able to help you do a Bode plot of this system, and define the stability margins. -- Dave Platt <dplatt(a)radagast.org> AE6EO Friends of Jade Warrior home page: http://www.radagast.org/jade-warrior I do _not_ wish to receive unsolicited commercial email, and I will boycott any company which has the gall to send me such ads!
From: Michael Robinson on 23 Dec 2009 14:17 "Dave Platt" <dplatt(a)radagast.org> wrote in message news:h32807-i82.ln1(a)radagast.org... > In article <AIbYm.104465$Wf2.40992(a)newsfe23.iad>, > Michael Robinson <nospam(a)billburg.com> wrote: > >>As I explained from the beginning, the votlage divider at the reference >>pin >>is connected to the battery/alternator output, like this: >> >> field >> | >> mosfet >> | >> vco batt >> | | >> +12v---[R]---+------+ | >> | | | >> __|__/ [C] [R] >> / / \ | | >> /___\----+-------+ >> | | >> | [R] >> | | >> Gnd ----------+--------------+ >> >>Can you see how it would regulate voltage? Any variation in the battery >>voltage will cause the cathode to pull the vco up or down, changing the >>duty >>cycle of the mosfet. It's a big feedback loop that includes the >>alternator. >>I want to know how the cap throws everything off. > > It seems to me that the presence of the cap will necessarily delay the > "visibility" of any change in the battery terminal voltage. The cap > will tend to hold the ref/C/R/R junction at a constant voltage, even > if the battery changes voltage rather abruptly and substantially. > > This means that the VCO's ability to react (increasing the MOSFET duty > cycle and the field) will be seriously compromised. > > It'll be slow in reacting (and increasing the field) when the battery > voltage drops. Similarly, it will be slow in reacting and decreasing > the field, when the battery voltage climbs too high. > > I think you can get (and have gotten!) yourself into trouble, in any > situation where the RC time constant created here, is slower than the > speed at which the field strength (and alternator output voltage) > reacts to changes in the VCO control voltage. The battery voltage > will tend to overshoot in both directions, because the control circuit > is *always* trying to catch up with (and react to) the overly-strong > reaction to its previous change. At best, the output will ring. At > worst, it will oscillate. > > This is not dissimilar to a classic stability problem in audio > amplifiers... trying to use a fast front end circuit, with slow > power-stage transistors. Using a fast-acting control system, with a > lot of gain, in a slow feedback loop, is a common recipe for serious > instability. > > I think that in order to ensure stability, you need to make sure that > the RC time constant involved isn't larger than the amount of time > required for the changes in VCO control voltage to take effect at the > battery. > > Another approach might be to reduce the gain of the control system... > i.e. the amount by which a change in the VCO control voltage affects > the duty cycle of the alternator. > > Someone with more EE knowledge than I, might be able to help you do a > Bode plot of this system, and define the stability margins. > > -- > Dave Platt <dplatt(a)radagast.org> AE6EO I'll have to study Bode plots. Thanks for the tip.
From: ehsjr on 23 Dec 2009 21:53
Michael Robinson wrote: <snip> >> >>Drawing it out will make it clear. The first schematic below >>is a "sanity check", just to make sure that the circuit >>you described is something like it: > > > As I explained from the beginning, the votlage divider at the reference pin > is connected to the battery/alternator output, like this: > > field > | > mosfet > | > vco batt > | | > +12v---[R]---+------+ | > | | | > __|__/ [C] [R] > / / \ | | > /___\----+-------+ > | | > | [R] > | | > Gnd ----------+--------------+ > > Can you see how it would regulate voltage? Yes. I've seen the idea all along, just not sure what was connected to where. Doesn't change my earlier answer. > Any variation in the battery > voltage will cause the cathode to pull the vco up or down, changing the duty > cycle of the mosfet. It's a big feedback loop that includes the alternator. > I want to know how the cap throws everything off. Check the diagram I drew showing the 431 as an op amp & output xsistor, with an internal 2.5 V ref. The 431 works as a comparator driving a transistor, and your cap feeds the output back into the op amp input, causing oscillation. Your TL431 will operate (turn on or off) in response to a voltage change at the battery. The current through R will suddenly change when the 431 switches on or off, changing the drop across R. *That* change will be fed back by the cap into the input of the 431 at the ref pin as a pulse, and the 431 will react to that pulse. That feedback path is causing your problem. You do not want *that* change (the change at the cathode of the TL431) to appear on the ref pin. Ed > > |