search

HSU-ANT / ACME.jl

ACME.jl - Analog Circuit Modeling and Emulation for Julia
Other
142 stars 24 forks source link

Triode / Pentode Models #29

Open ghost opened 4 years ago

ghost commented 4 years ago

@martinholters , I've take a look to the differents triode/pentode/... models used in the scientific litterature, I see you take part in differents study. For example the 2011 paper : "A Triode Model for Guitar Amplifier Simulation with Individual Parameter Fitting" with Kristjan Dempwolf and Udo Zölzer.

Can I ask you if you can orient me to the current state of the art for this subject? You may be advised on the subject. If not, no problem, I will take few weeks to work on the subject.

martinholters commented 4 years ago

Although I've been somewhat involved with Kristjan Dempwolfs work, I'm not sure about my qualification concerning the current state of the art. I'm not aware of any significant publications on the subject since then, but I may easily have missed some.

If I remember correctly, we were quite happy with our results on the triode, so I would probably have started with that model. But I'm obviously biased here :smile: We also did some work on pentodes, but as far as I remember, results were a bit mixed there. It might even make sense to implement different models to facilitate further research in this area. With the obvious downside that requires, well, implementing different models.

ghost commented 4 years ago

So I will take probably few days before make a little feedback with the idea, and maybe a little help from you to validate my work ;)

cristianfarias commented 3 years ago

Hi Martin,

I added the next triode element (COHEN AND HELIE model) but it seems to not resolve successfully the initial conditions ? 1) First I add the next code in the file elements.jl

export resistor, potentiometer, capacitor, inductor, transformer, voltagesource, currentsource, voltageprobe, currentprobe, diode, bjt, mosfet, opamp, triode

... ...

@doc raw""" triode(;Gk=2.14e-3, uk=100.8, Ek=1.303, Ck=3.04, Gg=6.06e-4, Eg=1.354, Cg=13.9)

Creates a triode model

Pins: plate, grid, cathode """ function triode(;Gk=2.14e-3, uk=100.8, Ek=1.303, Ck=3.04, Gg=6.06e-4, Eg=1.354, Cg=13.9)

nonlinear_eq =
    @inline function (q)
        VgkT1, VpkT1, IgT1, IpT1 = q
        IgT1=Gg*(log(1+exp(Cg*VgkT1))*1/Cg)^Eg
        IpT1=Gk*(log(1+exp(Ck*(VpkT1/uk+VgkT1)))*1/Ck)^Ek-Gg*(log(1+exp(Cg*VgkT1))*1/Cg)^Eg
        dIgT1_dVgkT1=(1891999529292056969*exp((139*VgkT1)/10)*((10*log(exp((139*VgkT1)/10) + 1))/139)^(177/500))/(2305843009213693952000*(exp((139*VgkT1)/10) + 1))
        dIgT1_dVpkT1=0
        dIpT1_dVgkT1=(139421*exp((76*VgkT1)/25 + (19*VpkT1)/630)*((25*log(exp((76*VgkT1)/25 + (19*VpkT1)/630) + 1))/76)^(303/1000))/(50000000*(exp((76*VgkT1)/25 + (19*VpkT1)/630) + 1)) - (1891999529292056969*exp((139*VgkT1)/10)*((10*log(exp((139*VgkT1)/10) + 1))/139)^(177/500))/(2305843009213693952000*(exp((139*VgkT1)/10) + 1))
        dIpT1_dVpkT1=(139421*exp((76*VgkT1)/25 + (19*VpkT1)/630)*((25*log(exp((76*VgkT1)/25 + (19*VpkT1)/630) + 1))/76)^(303/1000))/(5040000000*(exp((76*VgkT1)/25 + (19*VpkT1)/630) + 1))
        res = @SVector [IgT1, IpT1]
        J = @SMatrix [dIgT1_dVgkT1 dIgT1_dVpkT1 -1.0 0.0;
                      dIpT1_dVgkT1 dIpT1_dVpkT1 0.0  -1.0]
        return (res, J)
    end
return Element(mv=[1 0; 0 1; 0 0; 0 0],
               mi = [0 0; 0 0; 1 0; 0 1],
               mq = [-1 0 0 0; 0 -1 0 0; 0 0 -1 0; 0 0 0 -1], nonlinear_eq = nonlinear_eq,
               ports = [:grid => :cathode, :plate => :cathode])

end

2) When I tried to calculate the model I get the next error

julia> model = DiscreteModel(circ, 1/44100) ERROR: Failed to find initial solution Stacktrace: [1] error(::String) at .\error.jl:33 [2] initial_solution(::Function, ::Array{Float64,1}, ::Int64) at C:\Users\cti4121.julia\packages\ACME\bLRnf\src\ACME.jl:436 [3] DiscreteModel(::Circuit, ::Float64, ::Type{HomotopySolver{CachingSolver{SimpleSolver}}}; decompose_nonlinearity::Bool) at C:\Users\cti4121.julia\packages\ACME\bLRnf\src\ACME.jl:198 [4] DiscreteModel at C:\Users\cti4121.julia\packages\ACME\bLRnf\src\ACME.jl:150 [inlined] (repeats 2 times)

cristianfarias commented 3 years ago

before that, I define this circuit to simulate

vol=0.5;

circ = @circuit begin j3 = voltagesource(250), [-] ⟷ gnd, [+] ⟷ vcc # 250V power supply j1 = voltagesource(), [-] ⟷ gnd # input c1 = capacitor(47e-9), [1] ⟷ j1[+] r1 = resistor(1e6), [1] ⟷ c1[2], [2] ⟷ gnd r2 = resistor(70e3), [1] ⟷ c1[2] c2 = capacitor(2.5e-9), [1] ⟷ r2[2] t1 = triode(), [grid] ⟷ r2[2], [plate] ⟷ c2[2] r3 = resistor(100e3), [1] ⟷ vcc, [2] ⟷ t1[plate] r4 = resistor(1.5e3), [1] ⟷ t1[cathode], [2] ⟷ gnd c3 = capacitor(25e-6), [1] ⟷ t1[cathode], [2] ⟷ gnd c4 = capacitor(1e-6), [1] ⟷ t1[plate] p1 = potentiometer(100e3, (vol == nothing ? () : (vol,))...), [1] ⟷ gnd, [3] ⟷ c4[2] j2 = voltageprobe(), [-] ⟷ gnd, [+] ⟷ p1[2]# output end

Regards

martinholters commented 3 years ago

From a brief look, I think you'd want something like

res1 = Gg*(log(1+exp(Cg*VgkT1))*1/Cg)^Eg - IgT1
res2 = Gk*(log(1+exp(Ck*(VpkT1/uk+VgkT1)))*1/Ck)^Ek-Gg*(log(1+exp(Cg*VgkT1))*1/Cg)^Eg - IpT1
# ...
res = @SVector [res1, res]

I.e., the res vector has to hold the difference between the currents computed from the voltages and the currents provided as q[3] and q[4]. For convergence, it important that the Jacobian J is correct, which I cannot judge from a brief look, but given the strange numbers occurring in there, I assume you used some special tool to compute the derivatives for specific parameter values (probably the defaults)?

EDIT: I've just tried

function triode(;Gk=2.14e-3, uk=100.8, Ek=1.303, Ck=3.04, Gg=6.06e-4, Eg=1.354, Cg=13.9)
    nonlinear_eq =
        @inline function (q)
            VgkT1, VpkT1, IgT1, IpT1 = q
            res1 = Gg*(log(1+exp(Cg*VgkT1))*1/Cg)^Eg - IgT1
            res2 = Gk*(log(1+exp(Ck*(VpkT1/uk+VgkT1)))*1/Ck)^Ek-Gg*(log(1+exp(Cg*VgkT1))*1/Cg)^Eg - IpT1
                    dIgT1_dVgkT1=(1891999529292056969*exp((139*VgkT1)/10)*((10*log(exp((139*VgkT1)/10) + 1))/139)^(177/500))/(2305843009213693952000*(exp((139*VgkT1)/10) + 1))
                    dIgT1_dVpkT1=0
                    dIpT1_dVgkT1=(139421*exp((76*VgkT1)/25 + (19*VpkT1)/630)*((25*log(exp((76*VgkT1)/25 + (19*VpkT1)/630) + 1))/76)^(303/1000))/(50000000*(exp((76*VgkT1)/25 + (19*VpkT1)/630) + 1)) - (1891999529292056969*exp((139*VgkT1)/10)*((10*log(exp((139*VgkT1)/10) + 1))/139)^(177/500))/(2305843009213693952000*(exp((139*VgkT1)/10) + 1))
                    dIpT1_dVpkT1=(139421*exp((76*VgkT1)/25 + (19*VpkT1)/630)*((25*log(exp((76*VgkT1)/25 + (19*VpkT1)/630) + 1))/76)^(303/1000))/(5040000000*(exp((76*VgkT1)/25 + (19*VpkT1)/630) + 1))
            res = @SVector [res1, res2]
            J = @SMatrix [dIgT1_dVgkT1 dIgT1_dVpkT1 -1.0 0.0;
                          dIpT1_dVgkT1 dIpT1_dVpkT1 0.0  -1.0]
            return (res, J)
        end
    return ACME.Element(mv=[1 0; 0 1; 0 0; 0 0],
                   mi = [0 0; 0 0; 1 0; 0 1],
                   mq = [-1 0 0 0; 0 -1 0 0; 0 0 -1 0; 0 0 0 -1], nonlinear_eq = nonlinear_eq,
                   ports = [:grid => :cathode, :plate => :cathode])
end

With the circuit you gave above it a) doesn't error and b) gives results that look plausible (at first glance, at least).

cristianfarias commented 3 years ago

Thanks Martin for answer my question !!

After I write you, I make another changes to de file "elements.js" that made it works succefully the triode model that I was trying to simulated ( the code had a mistake that a mark below in yellow colour, and also I put the equations in symbolic mode now, and also has two types of triode models)

I resume all the changes (And I attach the new file "elements.jl")

1) First in the begining of the file "elements.js" I add the new 2 models, one for KOHEN AND HELLIE model of triode, one for KOREN model of triode

export resistor, potentiometer, capacitor, inductor, transformer, voltagesource, currentsource, voltageprobe, currentprobe, diode, bjt, mosfet, opamp, triode_koh, triode_kor

2) In the end of file "elements.js" I add the 2 new models

@doc raw""" triode_koh(;Gk=2.14e-3, uk=100.8, Ek=1.303, Ck=3.04, Gg=6.06e-4, Eg=1.354, Cg=13.9)

Creates a triode KOHEN AND HELIE model

Pins: plate, grid, cathode """ function triode_koh(;Gk=2.14e-3, uk=100.8, Ek=1.303, Ck=3.04, Gg=6.06e-4, Eg=1.354, Cg=13.9)

nonlinear_eq =
    @inline function (q)
        VgkT1, VpkT1, Ig, Ip = q
        IgT1=Gg*(log(1+exp(Cg*VgkT1))*1/Cg)^Eg
        IpT1=Gk*(log(1+exp(Ck*(VpkT1/uk+VgkT1)))*1/Ck)^Ek-Gg*(log(1+exp(Cg*VgkT1))*1/Cg)^Eg

dIgT1_dVgkT1=(EgGgexp(CgVgkT1)(log(exp(CgVgkT1) + 1)/Cg)^(Eg - 1))/(exp(CgVgkT1) + 1) dIgT1_dVpkT1=0 dIpT1_dVgkT1=(EkGkexp(Ck(VgkT1 + VpkT1/uk))(log(exp(Ck(VgkT1 + VpkT1/uk)) + 1)/Ck)^(Ek - 1))/(exp(Ck(VgkT1 + VpkT1/uk)) + 1) - (EgGgexp(CgVgkT1)(log(exp(CgVgkT1) + 1)/Cg)^(Eg - 1))/(exp(CgVgkT1) + 1) dIpT1_dVpkT1=(EkGkexp(Ck(VgkT1 + VpkT1/uk))(log(exp(Ck(VgkT1 + VpkT1/uk)) + 1)/Ck)^(Ek - 1))/(uk(exp(Ck*(VgkT1 + VpkT1/uk)) + 1)) res = @SVector [IgT1 - Ig, IpT1 - Ip] J = @SMatrix [dIgT1_dVgkT1 dIgT1_dVpkT1 -1.0 0.0; dIpT1_dVgkT1 dIpT1_dVpkT1 0.0 -1.0] return (res, J) end return Element(mv=[1 0; 0 1; 0 0; 0 0], mi = [0 0; 0 0; 1 0; 0 1], mq = [-1 0 0 0; 0 -1 0 0; 0 0 -1 0; 0 0 0 -1], nonlinear_eq = nonlinear_eq, ports = [:grid => :cathode, :plate => :cathode]) end

@doc raw""" triode_kor(;u=100, KX=1.4, KG1=1060, KP=600, KVB=300, VT=0.026, RG=2000)

Creates a triode KOREN model

Pins: plate, grid, cathode """ function triode_kor(;u=100, KX=1.4, KG1=1060, KP=600, KVB=300, VT=0.026, RG=2000)

nonlinear_eq =
    @inline function (q)
        VgkT1, VpkT1, Ig, Ip = q

dirac(n) = n == 0 IgT1=log(1+exp(VgkT1/VT))VT/RG Ed1=VpkT1/KPlog(1+exp(KP(1/u+VgkT1/sqrt(KVB+VpkT1^2)))) IpT1=Ed1^KX/KG1(1+sign(Ed1)) dIgT1_dVgkT1=exp(VgkT1/VT)/(RG(exp(VgkT1/VT) + 1)) dIgT1_dVpkT1=0 dIpT1_dVgkT1=(2VpkT1dirac((VpkT1log(exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)) + 1))/KP)exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u))((VpkT1log(exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)) + 1))/KP)^KX)/(KG1(exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)) + 1)(VpkT1^2 + KVB)^(1/2)) + (KXVpkT1exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u))(sign((VpkT1log(exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)) + 1))/KP) + 1)((VpkT1log(exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)) + 1))/KP)^(KX - 1))/(KG1(exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)) + 1)(VpkT1^2 + KVB)^(1/2)) dIpT1_dVpkT1=(2dirac((VpkT1log(exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)) + 1))/KP)(log(exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)) + 1)/KP - (VgkT1VpkT1^2exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)))/((exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)) + 1)(VpkT1^2 + KVB)^(3/2)))((VpkT1log(exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)) + 1))/KP)^KX)/KG1 + (KX(log(exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)) + 1)/KP - (VgkT1VpkT1^2exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)))/((exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)) + 1)(VpkT1^2 + KVB)^(3/2)))(sign((VpkT1log(exp(KP(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)) + 1))/KP) + 1)((VpkT1log(exp(KP*(VgkT1/(VpkT1^2 + KVB)^(1/2) + 1/u)) + 1))/KP)^(KX - 1))/KG1 res = @SVector [IgT1 - Ig, IpT1 - Ip] J = @SMatrix [dIgT1_dVgkT1 dIgT1_dVpkT1 -1.0 0.0; dIpT1_dVgkT1 dIpT1_dVpkT1 0.0 -1.0] return (res, J) end return Element(mv=[1 0; 0 1; 0 0; 0 0], mi = [0 0; 0 0; 1 0; 0 1], mq = [-1 0 0 0; 0 -1 0 0; 0 0 -1 0; 0 0 0 -1], nonlinear_eq = nonlinear_eq, ports = [:grid => :cathode, :plate => :cathode]) end

Both models works succefully to simulate the next circuit for example "Common Cathode Triode Amplifier"

vol=0.1;

circ = @circuit begin j3 = voltagesource(250), [-] ⟷ gnd, [+] ⟷ vcc # 250V power supply j1 = voltagesource(), [-] ⟷ gnd # input c1 = capacitor(47e-9), [1] ⟷ j1[+] r1 = resistor(1e6), [1] ⟷ c1[2], [2] ⟷ gnd r2 = resistor(70e3), [1] ⟷ c1[2] c2 = capacitor(2.5e-12), [1] ⟷ r2[2] t1 = triode_koh(), [grid] ⟷ r2[2], [plate] ⟷ c2[2] r3 = resistor(100e3), [1] ⟷ vcc, [2] ⟷ t1[plate] r4 = resistor(1.5e3), [1] ⟷ t1[cathode], [2] ⟷ gnd c3 = capacitor(25e-6), [1] ⟷ t1[cathode], [2] ⟷ gnd c4 = capacitor(1e-6), [1] ⟷ t1[plate] p1 = potentiometer(100e3, (vol == nothing ? () : (vol,))...), [1] ⟷ gnd, [3] ⟷ c4[2] j2 = voltageprobe(), [-] ⟷ gnd, [+] ⟷ p1[2]# output end

[cid:5a3dae34-68c4-464f-8e70-7c0d849891e4]

But when I try to simulate the next circuito with two 12AX7 triodes known as "Fender Type Guitar Preamp" the output isn't not the right one, and I get some error when calculate de model

[cid:4235bf6e-2adb-4179-9929-be0c956e9843] [cid:eba3e6fe-88c2-490d-b8e1-a93d2c5b5812]

vol6=0.5; vol7=0.5; vol8=0.5; vol9=0.5; vol10=0.1;

circ = @circuit begin j3 = voltagesource(385), [-] ⟷ gnd, [+] ⟷ vcc # 385V power supply j1 = voltagesource(), [-] ⟷ gnd # input r1 = resistor(32e3), [1] ⟷ j1[+] r2 = resistor(1e6), [1] ⟷ r1[2], [2] ⟷ gnd t1 = triode_kor(;u=100, KX=1.4, KG1=1060, KP=600, KVB=300, VT=0.026, RG=2000), [grid] ⟷ r1[2] r3 = resistor(1.5e3), [1] ⟷ t1[cathode], [2] ⟷ gnd c1 = capacitor(22e-6), [1] ⟷ t1[cathode], [2] ⟷ gnd r4 = resistor(100e3), [1] ⟷ vcc, [2] ⟷ t1[plate] r5 = resistor(100e3), [1] ⟷ t1[plate] c2 = capacitor(250e-12), [1] ⟷ t1[plate] c3 = capacitor(100e-9), [1] ⟷ r5[2] c4 = capacitor(22e-9), [1] ⟷ r5[2] p6 = potentiometer(250e3, (vol6 == nothing ? () : (vol6,))...), [1] ⟷ c3[2], [3] ⟷ c2[2] p7 = potentiometer(250e3, (vol7 == nothing ? () : (vol7,))...), [1] ⟷ c4[2], [3] ⟷ c3[2] p8 = potentiometer(10e3, (vol8 == nothing ? () : (vol8,))...), [1] ⟷ gnd, [3] ⟷ c4[2] p9 = potentiometer(1e6, (vol9 == nothing ? () : (vol9,))...), [1] ⟷ gnd, [3] ⟷ p6[2] c5 = capacitor(120e-12), [1] ⟷ p6[2], [2] ⟷ p9[2] t2 = triode_kor(;u=100, KX=1.4, KG1=1060, KP=600, KVB=300, VT=0.026, RG=2000), [grid] ⟷ p9[2] r10 = resistor(1640), [1] ⟷ t2[cathode], [2] ⟷ gnd c6 = capacitor(22e-6), [1] ⟷ t2[cathode], [2] ⟷ gnd r4 = resistor(100e3), [1] ⟷ vcc, [2] ⟷ t2[plate] c7 = capacitor(1e-6), [1] ⟷ t2[plate] p10 = potentiometer(400e3, (vol10 == nothing ? () : (vol10,))...), [1] ⟷ gnd, [3] ⟷ c7[2] j2 = voltageprobe(), [-] ⟷ gnd, [+] ⟷ p10[2]# output end

connect!(circ, (:p7, :2), (:p7, :3)) connect!(circ, (:p8, :2), (:p8, :3))

julia> model = DiscreteModel(circ, 1/44100) ERROR: DomainError with -6.035191317003306e-5: Exponentiation yielding a complex result requires a complex argument. Replace x^y with (x+0im)^y, Complex(x)^y, or similar.

However it allows me to graph the output, but no show distortion

y = run!(model, [0.5sin(2π1000/44100n) for c in 1:1, n in 0:44099]) in=[0.5sin(2π1000/44100n) for c in 1:1, n in 0:44099] x=[n for c in 1:1, n in 0:44099] z=transpose(y) plot(x[40000:40400],z[40000:40400]) plot!(x[40000:40400],in[40000:40400])

[cid:e2175b3c-83e8-4f9d-a1bd-7ce3257b442c]

De: Martin Holters notifications@github.com Enviado: martes, 9 de marzo de 2021 10:06 Para: HSU-ANT/ACME.jl ACME.jl@noreply.github.com Cc: cristianfarias cristianfarias@hotmail.com; Comment comment@noreply.github.com Asunto: Re: [HSU-ANT/ACME.jl] Triode / Pentode Models (#29)

From a brief look, I think you'd want something like

res1 = Gg(log(1+exp(CgVgkT1))1/Cg)^Eg - IgT1 res2 = Gk(log(1+exp(Ck(VpkT1/uk+VgkT1)))1/Ck)^Ek-Gg(log(1+exp(CgVgkT1))*1/Cg)^Eg - IpT1

...

res = @SVector [res1, res]

I.e., the res vector has to hold the difference between the currents computed from the voltages and the currents provided as q[3] and q[4]. For convergence, it important that the Jacobian J is correct, which I cannot judge from a brief look, but given the strange numbers occurring in there, I assume you used some special tool to compute the derivatives for specific parameter values (probably the defaults)?

— You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/HSU-ANT/ACME.jl/issues/29#issuecomment-793873272, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AEOH6L7ILVDA7WQXXLMIWUTTCYMMJANCNFSM4NM4C3NA.

cristianfarias commented 3 years ago

Hi Martin,

For example, the original equations than formulate the currents in the KOHEN AND HELIE model of a triode are the next:

IgT1=Gg(log(1+exp(CgVgkT1))*1/Cg)^Eg

IpT1=Gk(log(1+exp(Ck(VpkT1/uk+VgkT1)))1/Ck)^Ek-Gg(log(1+exp(CgVgkT1))1/Cg)^Eg

Using MATLAB I calculate the first derivatives from IgT1 and IpT1 respect to Vgk and Vpk , respectively

that I named in my code as follow ---- > dIgT1_dVgkT1 dIgT1_dVpkT1 dIpT1_dVgkT1 dIpT1_dVpkT1

Here are the command that I use in MATLAB to calculate them

syms VgkT1 VpkT1 Gg Cg Eg Gk Ck uk Ek; IgT1=Gg(log(1+exp(CgVgkT1))1/Cg)^Eg; IpT1=Gk(log(1+exp(Ck(VpkT1/uk+VgkT1)))1/Ck)^Ek-Gg(log(1+exp(CgVgkT1))*1/Cg)^Eg; diff(IgT1,VgkT1)

ans =

(Eg*Gg*exp(Cg*VgkT1)*(log(exp(Cg*VgkT1) + 1)/Cg)^(Eg - 1))/(exp(Cg*VgkT1) + 1)

diff(IgT1,VpkT1)

ans =

0

diff(IpT1,VgkT1)

ans =

(Ek*Gk*exp(Ck*(VgkT1 + VpkT1/uk))*(log(exp(Ck*(VgkT1 + VpkT1/uk)) + 1)/Ck)^(Ek - 1))/(exp(Ck*(VgkT1 + VpkT1/uk)) + 1) - (Eg*Gg*exp(Cg*VgkT1)*(log(exp(Cg*VgkT1) + 1)/Cg)^(Eg - 1))/(exp(Cg*VgkT1) + 1)

diff(IpT1,VpkT1)

ans =

(Ek*Gk*exp(Ck*(VgkT1 + VpkT1/uk))*(log(exp(Ck*(VgkT1 + VpkT1/uk)) + 1)/Ck)^(Ek - 1))/(uk*(exp(Ck*(VgkT1 + VpkT1/uk)) + 1))

Now, when a tried to simulate the circuit with two 12AX7 triodes known as "Fender Type Guitar Preamp", and using the KOHEN AND HELLIE model ----- > then there isn't any error in the commands, but the simulated output there isn't to be the expected ???

vol6=0.5; vol7=0.5; vol8=0.5; vol9=0.5; vol10=0.1;

circ = @circuit begin j3 = voltagesource(385), [-] ⟷ gnd, [+] ⟷ vcc # 385V power supply j1 = voltagesource(), [-] ⟷ gnd # input r1 = resistor(32e3), [1] ⟷ j1[+] r2 = resistor(1e6), [1] ⟷ r1[2], [2] ⟷ gnd t1 = triode_koh(;Gk=2.14e-3, uk=100.8, Ek=1.303, Ck=3.04, Gg=6.06e-4, Eg=1.354, Cg=13.9), [grid] ⟷ r1[2] r3 = resistor(1.5e3), [1] ⟷ t1[cathode], [2] ⟷ gnd c1 = capacitor(22e-6), [1] ⟷ t1[cathode], [2] ⟷ gnd r4 = resistor(100e3), [1] ⟷ vcc, [2] ⟷ t1[plate] r5 = resistor(100e3), [1] ⟷ t1[plate] c2 = capacitor(250e-12), [1] ⟷ t1[plate] c3 = capacitor(100e-9), [1] ⟷ r5[2] c4 = capacitor(22e-9), [1] ⟷ r5[2] p6 = potentiometer(250e3, (vol6 == nothing ? () : (vol6,))...), [1] ⟷ c3[2], [3] ⟷ c2[2] p7 = potentiometer(250e3, (vol7 == nothing ? () : (vol7,))...), [1] ⟷ c4[2], [3] ⟷ c3[2] p8 = potentiometer(10e3, (vol8 == nothing ? () : (vol8,))...), [1] ⟷ gnd, [3] ⟷ c4[2] p9 = potentiometer(1e6, (vol9 == nothing ? () : (vol9,))...), [1] ⟷ gnd, [3] ⟷ p6[2] c5 = capacitor(120e-12), [1] ⟷ p6[2], [2] ⟷ p9[2] t2 = triode_koh(;Gk=2.14e-3, uk=100.8, Ek=1.303, Ck=3.04, Gg=6.06e-4, Eg=1.354, Cg=13.9), [grid] ⟷ p9[2] r10 = resistor(1640), [1] ⟷ t2[cathode], [2] ⟷ gnd c6 = capacitor(22e-6), [1] ⟷ t2[cathode], [2] ⟷ gnd r4 = resistor(100e3), [1] ⟷ vcc, [2] ⟷ t2[plate] c7 = capacitor(1e-6), [1] ⟷ t2[plate] p10 = potentiometer(400e3, (vol10 == nothing ? () : (vol10,))...), [1] ⟷ gnd, [3] ⟷ c7[2] j2 = voltageprobe(), [-] ⟷ gnd, [+] ⟷ p10[2]# output end

connect!(circ, (:p7, :2), (:p7, :3)) connect!(circ, (:p8, :2), (:p8, :3))

model = DiscreteModel(circ, 1/44100)

y = run!(model, [0.5sin(2π1000/44100n) for c in 1:1, n in 0:44099]) in=[0.5sin(2π1000/44100n) for c in 1:1, n in 0:44099] x=[n for c in 1:1, n in 0:44099] z=transpose(y) plot(x[40000:40400],z[40000:40400]) plot!(x[40000:40400],in[40000:40400])

Regards !!

De: Martin Holters notifications@github.com Enviado: martes, 9 de marzo de 2021 10:06 Para: HSU-ANT/ACME.jl ACME.jl@noreply.github.com Cc: cristianfarias cristianfarias@hotmail.com; Comment comment@noreply.github.com Asunto: Re: [HSU-ANT/ACME.jl] Triode / Pentode Models (#29)

From a brief look, I think you'd want something like

res1 = Gg(log(1+exp(CgVgkT1))1/Cg)^Eg - IgT1 res2 = Gk(log(1+exp(Ck(VpkT1/uk+VgkT1)))1/Ck)^Ek-Gg(log(1+exp(CgVgkT1))*1/Cg)^Eg - IpT1

...

res = @SVector [res1, res]

I.e., the res vector has to hold the difference between the currents computed from the voltages and the currents provided as q[3] and q[4]. For convergence, it important that the Jacobian J is correct, which I cannot judge from a brief look, but given the strange numbers occurring in there, I assume you used some special tool to compute the derivatives for specific parameter values (probably the defaults)?

— You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/HSU-ANT/ACME.jl/issues/29#issuecomment-793873272, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AEOH6L7ILVDA7WQXXLMIWUTTCYMMJANCNFSM4NM4C3NA.

martinholters commented 3 years ago

Sorry, your latest posts somehow slipped through unnoticed. Probably a bit late now, but in your example circuit, there are two r4's. Renaming the latter to, say, r11, yields plausible results. (Using the same designator twice should probably give a warning to help avoid such pitfalls).

cristianfarias commented 3 years ago

Thanks Martin for your help.

I rename this variable and the output was the expected !!!

Regards

De: Martin Holters @.> Enviado: martes, 20 de julio de 2021 03:04 Para: HSU-ANT/ACME.jl @.> Cc: cristianfarias @.>; Comment @.> Asunto: Re: [HSU-ANT/ACME.jl] Triode / Pentode Models (#29)

Sorry, your latest posts somehow slipped through unnoticed. Probably a bit late now, but in your example circuit, there are two r4's. Renaming the latter to, say, r11, yields plausible results. (Using the same designator twice should probably give a warning to help avoid such pitfalls).

— You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/HSU-ANT/ACME.jl/issues/29#issuecomment-883096131, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AEOH6L53XFSFX32XJQVFLNDTYUGWLANCNFSM4NM4C3NA.

martinholters commented 3 years ago

FWIW, with the recently released v0.9.5, the buggy circuit description above now emits

┌ Warning: redefinition of `r4`