1. Forum
  2. Usenet
  3. SCI.ELECTRONICS.DESIGN

  • dumper circuit

    From john larkin@21:1/5 to All on Sat Mar 8 08:57:43 2025
    I'll have an isolated class-D amplifier powered by a dc/dc converter
    from my big 48 volt supply. Simple push-pull, transformer, rectifier,
    making maybe isolated 46.

    In one situation, the class-D amp can push current uphill into the
    isolated supply, and we need to dump it. This circuit acts sort of
    like a 25-watt zener diode.

    https://www.dropbox.com/scl/fi/98n07mnf3qokvbewuv1oy/Dumper_1.jpg?rlkey=poiep8hzhz33qecswdj50sdne&raw=1

    It didn't work in simulation. It was driving me to despair. No amount
    of chocolate would help.

    The problem was of course that the Spice initial condition simulation
    perfectly biased everything, balanced the pencil on its point.

    So how can the initial conditions sim ignore the positive feedback?

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From john larkin @21:1/5 to sunaecoNoChoppedPork@gmail.com on Sat Mar 8 12:03:43 2025
    On Sat, 08 Mar 2025 18:57:11 +0000, JM
    <sunaecoNoChoppedPork@gmail.com> wrote:

    On Sat, 08 Mar 2025 08:57:43 -0800, john larkin <jl@650pot.com> wrote:

    I'll have an isolated class-D amplifier powered by a dc/dc converter
    from my big 48 volt supply. Simple push-pull, transformer, rectifier, >>making maybe isolated 46.

    In one situation, the class-D amp can push current uphill into the
    isolated supply, and we need to dump it. This circuit acts sort of
    like a 25-watt zener diode.
    https://www.dropbox.com/scl/fi/98n07mnf3qokvbewuv1oy/Dumper_1.jpg?rlkey=poiep8hzhz33qecswdj50sdne&raw=1

    It didn't work in simulation. It was driving me to despair. No amount
    of chocolate would help.

    The problem was of course that the Spice initial condition simulation >>perfectly biased everything, balanced the pencil on its point.

    So how can the initial conditions sim ignore the positive feedback?

    Use .ic together with uic.

    It sims fine if I click the box to skip the initial condition
    solution. That's faster too. Delaying the power supplies breaks the
    tie too.

    What I don't understand is how negative feedbacks work during the ic
    solution, but positive feedbacks don't.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From john larkin @21:1/5 to bitrex on Sat Mar 8 14:06:50 2025
    On Sat, 8 Mar 2025 16:46:14 -0500, bitrex <user@example.net> wrote:

    On 3/8/2025 3:03 PM, john larkin wrote:
    On Sat, 08 Mar 2025 18:57:11 +0000, JM
    <sunaecoNoChoppedPork@gmail.com> wrote:

    On Sat, 08 Mar 2025 08:57:43 -0800, john larkin <jl@650pot.com> wrote:

    I'll have an isolated class-D amplifier powered by a dc/dc converter
    from my big 48 volt supply. Simple push-pull, transformer, rectifier,
    making maybe isolated 46.

    In one situation, the class-D amp can push current uphill into the
    isolated supply, and we need to dump it. This circuit acts sort of
    like a 25-watt zener diode.

    https://www.dropbox.com/scl/fi/98n07mnf3qokvbewuv1oy/Dumper_1.jpg?rlkey=poiep8hzhz33qecswdj50sdne&raw=1

    It didn't work in simulation. It was driving me to despair. No amount
    of chocolate would help.

    The problem was of course that the Spice initial condition simulation
    perfectly biased everything, balanced the pencil on its point.

    So how can the initial conditions sim ignore the positive feedback?

    Use .ic together with uic.

    It sims fine if I click the box to skip the initial condition
    solution. That's faster too. Delaying the power supplies breaks the
    tie too.

    What I don't understand is how negative feedbacks work during the ic
    solution, but positive feedbacks don't.



    Strongly non-linear ODEs and systems of ODEs are more likely to have
    singular solutions i.e. parts of the domain where the initial value
    problem fails to return a unique answer...I'm not an expert in how this >translates to discretized/numerical solutions but I believe the same
    sort of thing applies, a system of finite difference equations
    describing a positive-feedback circuit is more likely to be strongly >non-linear than one for negative feedback.

    Sometimes an initial value problem is only troublesome at a single point
    like t = 0 and if you move off that point the problem is well-posed.

    Version 4
    SHEET 1 880 680
    WIRE 272 -16 208 -16
    WIRE 336 -16 272 -16
    WIRE 336 0 336 -16
    WIRE 336 96 336 80
    WIRE 208 112 208 -16
    WIRE 80 128 32 128
    WIRE 176 128 80 128
    WIRE 272 144 240 144
    WIRE 288 144 272 144
    WIRE 304 144 288 144
    WIRE 176 160 128 160
    WIRE 208 208 208 176
    WIRE 128 240 128 160
    WIRE 272 240 272 144
    WIRE 272 240 128 240
    WIRE 32 288 32 128
    WIRE 208 288 32 288
    WIRE 272 288 208 288
    WIRE 32 320 32 288
    WIRE 272 320 272 288
    WIRE 32 432 32 400
    WIRE 224 448 208 448
    WIRE 272 448 272 400
    WIRE 272 448 224 448
    FLAG 208 208 0
    FLAG 336 96 0
    FLAG 32 432 0
    FLAG 272 -16 +10
    FLAG 224 448 +10
    FLAG 80 128 N
    FLAG 208 288 P
    FLAG 288 144 U
    SYMBOL OpAmps\\UniversalOpAmp2 208 144 R0
    WINDOW 0 -63 48 Left 2
    SYMATTR InstName U1
    SYMBOL voltage 336 -16 R0
    SYMATTR InstName V1
    SYMATTR Value 10
    SYMBOL res 48 416 R180
    WINDOW 0 -52 67 Left 2
    WINDOW 3 -58 38 Left 2
    SYMATTR InstName R2
    SYMATTR Value 10K
    SYMBOL res 288 416 R180
    WINDOW 0 -46 69 Left 2
    WINDOW 3 -54 39 Left 2
    SYMATTR InstName R4
    SYMATTR Value 10K
    TEXT -176 304 Left 2 !.tran 1
    TEXT -224 208 Left 2 ;Opamp Startup
    TEXT -224 248 Left 2 ;JL Mar 8 2025


    Run that for 1 second and look at node U.

    Repeat but run for 10 seconds.

    Then run for 1000 seconds.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From bitrex@21:1/5 to john larkin on Sat Mar 8 16:46:14 2025
    On 3/8/2025 3:03 PM, john larkin wrote:
    On Sat, 08 Mar 2025 18:57:11 +0000, JM
    <sunaecoNoChoppedPork@gmail.com> wrote:

    On Sat, 08 Mar 2025 08:57:43 -0800, john larkin <jl@650pot.com> wrote:

    I'll have an isolated class-D amplifier powered by a dc/dc converter
    from my big 48 volt supply. Simple push-pull, transformer, rectifier,
    making maybe isolated 46.

    In one situation, the class-D amp can push current uphill into the
    isolated supply, and we need to dump it. This circuit acts sort of
    like a 25-watt zener diode.

    https://www.dropbox.com/scl/fi/98n07mnf3qokvbewuv1oy/Dumper_1.jpg?rlkey=poiep8hzhz33qecswdj50sdne&raw=1

    It didn't work in simulation. It was driving me to despair. No amount
    of chocolate would help.

    The problem was of course that the Spice initial condition simulation
    perfectly biased everything, balanced the pencil on its point.

    So how can the initial conditions sim ignore the positive feedback?

    Use .ic together with uic.

    It sims fine if I click the box to skip the initial condition
    solution. That's faster too. Delaying the power supplies breaks the
    tie too.

    What I don't understand is how negative feedbacks work during the ic solution, but positive feedbacks don't.



    Strongly non-linear ODEs and systems of ODEs are more likely to have
    singular solutions i.e. parts of the domain where the initial value
    problem fails to return a unique answer...I'm not an expert in how this translates to discretized/numerical solutions but I believe the same
    sort of thing applies, a system of finite difference equations
    describing a positive-feedback circuit is more likely to be strongly
    non-linear than one for negative feedback.

    Sometimes an initial value problem is only troublesome at a single point
    like t = 0 and if you move off that point the problem is well-posed.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)