OTL 2013

[audiodesign]

Freeware LT Spice IV circuit simulator

I am using the freeware LT Spice IV circuit simulator:

Linear Technology - Design Simulation and Device Models

all my simulation files are stored on:

Index of /otl/Sim

in same case it is necessary overwrite the original library in:

c:\program files\LTC\lib\cmp

to add some transistors, jfet and mosfet.

I can help you to use this program to make the different simulations: Transient and AC analysis.

 

[cnpope]

I am using the freeware LT Spice IV circuit simulator:

Linear Technology - Design Simulation and Device Models

all my simulation files are stored on:

Index of /otl/Sim

in same case it is necessary overwrite the original library in:

c:\program files\LTC\lib\cmp

to add some transistors, jfet and mosfet.

I can help you to use this program to make the different simulations: Transient and AC analysis.

Thanks! I'll try playing with these at the weekend, and get back to you if (probably, when) I have questions...

Chris

 

[jazbo8]

Which version of the 6C33C model are you using in the simulations?

 

[audiodesign]

Which version of the 6C33C model are you using in the simulations?

.subckt 6c33c p g k ;Rydel's Models
+params: g=2.957e-3 B=16.976 mu=1.984
+ k=0.941 Vc=-0.578
+ cgk=11p cgp=5p
+ cpk=1.8p
e1 A1 0 value= {1+(v(g,k)/(B-(V(g,k)/k)))}
re1 A1 0 100meg
e2 A2 0 value= {v(g,k)+((v(p,k)+Vc)/mu)}
re2 A2 0 100meg
g1 p k value= {g*v(A1)*(pwr(v(A2),1.5)+pwrs(v(A2),1.5))}
rpk p k 100meg
c1 g k {cgk}
c2 g p {cgp}
c3 p k {cpk}
.ends

 

[jazbo8]

6C33C Rydel model vs. datasheet. It does not appear to be so accurate and needs some adjustments.

Plate Characteristic

An externally hosted image should be here but it was not working when we last tested it.

Transfer Characteristic with Vp=120V (Zoom view)

An externally hosted image should be here but it was not working when we last tested it.

 

[audiodesign]

I have found these models:

.subckt 6c33c p g k ;Rydel's Models
+params: g=2.957e-3 B=16.976 mu=1.984
+ k=0.941 Vc=-0.578
+ cgk=11p cgp=5p
+ cpk=1.8p
e1 A1 0 value= {1+(v(g,k)/(B-(V(g,k)/k)))}
re1 A1 0 100meg
e2 A2 0 value= {v(g,k)+((v(p,k)+Vc)/mu)}
re2 A2 0 100meg
g1 p k value= {g*v(A1)*(pwr(v(A2),1.5)+pwrs(v(A2),1.5))}
rpk p k 100meg
c1 g k {cgk}
c2 g p {cgp}
c3 p k {cpk}
.ends

.subckt 6c33c_1 p g k ;Rydel's Models
+params: g=2.957e-3 B=16.976 mu=1.984
+ k=0.941 Vc=-0.578
+ cgk=11p cgp=5p
+ cpk=1.8p
e1 1 0 value= {1+(v(g,k)/(B-(V(g,k)/k)))}
re1 1 0 100meg
e2 2 0 value= {v(g,k)+((v(p,k)+Vc)/mu)}
re2 2 0 100meg
g1 p k value= {g*v(1)*(pwr(v(2),1.5)+pwrs(v(2),1.5))}
rpk p k 100meg
c1 g k {cgk}
c2 g p {cgp}
c3 p k {cpk}
.ends

.subckt 6c33c_2 p g k ;Rydel's Models
+params: g=2.957e-3 B=16.976 mu=2.2
+ k=0.941 Vc=-0.578
+ cgk=11p cgp=5p
+ cpk=1.8p
v_eddy 10 0 2.984
R_eddy1 10 11 0.504k
R-break 11 0 {m}
e1 1 0 value= {1+(v(g,k)/(B-(V(g,k)/k)))}
re1 1 0 100meg
e2 2 0 value= {v(g,k)+((v(p,k)+Vc)/v(11))}
re2 2 0 100meg
g1 p k value= {g*v(1)*(pwr(v(2),1.5)+pwrs(v(2),1.5))}
rpk p k 100meg
c1 g k {cgk}
c2 g p {cgp}
c3 p k {cpk}
.ends 6c33c

.SUBCKT 6C33C_OLD 1 2 3 ; A G C (Triode) ; TWO CATHODES FROM BORBELY, GA 5/96.
* Original Koren Lib
X1 1 2 3 TRIODE MU=3.1 EX=1.4 KG1=163 KP=15 KVB=300 VCT=0.00 RGI=1000 CCG=30.0P CGP=31.0P CCP=10.5P ; ADD .7PF TO ADJACENT PINS; .5 TO OTHERS.
.ENDS 6C33C

******************
.SUBCKT 6C33C1 1 2 3 ; P G C (Triode)
* Soviet (1970) Using both cathodes ie both heaters working
* library format: LTSpice 02-Jun-2008
X1 1 2 3 TRIODE MU=2.7 EX=1.397 KG1=150.9 KP=13.94 KVB=300.0 VCT=0.00 RGI=1000 CCG=30.0p CGP=31.0p CCP=10.5p ;
.ENDS 6C33C

 

[jazbo8]

Ran some more checks on the SPICE models, the 12AU7N was spot on, but the D3A did poorly and could use some adjustments. I admire your in-depth modeling efforts and meticulous documentation, but instead of relying on old SPICE models of questionable origins, perhaps the results could be improved further if you build them from scratch using tools like CurveCaptor or PaintModelTool. 😉

Here is another example of a questionable SPICE model. D3A model vs. datasheet. Whoa!

Transfer Characteristic (Vp=Vg2=150V)

An externally hosted image should be here but it was not working when we last tested it.

 

[audiodesign]

Yes, simulations is not the reality but this help to find the correct way.

Here an example of good result obtain with spice models:

6j51p | Bartola Valves

 

[jazbo8]

Yes, simulations is not the reality but this help to find the correct way.

Here an example of good result obtain with spice models:

6j51p | Bartola Valves

Yes, Ale is going one step further and getting very good results using curves from the actual tubes (not datasheets) and Dmitry´s model paint tool.

 

[tubetvr]

Spice modeling of 6C33C is difficult as it has a kink at low currents that doesn't exist on many other tubes. The kink affects distortion figures for 3rd order and higher products and it is essential to include this in the spice model , (if the model should be correct for distortion) I have tested most available spice models for 6C33C and they goes from barely acceptable to terrible and you still get large differences in results when comparing with a real amplifier.

It doesn't help either when the available characteristics curves that exist for 6C33C differ quite a lot, some have the kink clearly included and some have it just indicated but smoothed out which doesn't fit with reality, no available curves goes to several amperes of anode current either which is unfortunate as this is where the tube operates when used in an OTL.

One of the better available spice models for 6C33C I have found is from here http://www.excem.fr/download/usergui5.pdf, it has a correction for the kink but it is not exact.

 

[jazbo8]

Thanks for the tip, I did not realize how difficult it was to make an accurate model for the 6C33C. In addition to the Excem model, I also found Dmitry´s model here (alas at 6.3V filament only), based on actual tube curves from Audiomatica. Will try to do some comparisons of both models when I get a chance, should be interesting...

 

[jazbo8]

Upon further reading - the Excem model is the same as the Rydel model (with some adjustments to the parameters), so as you said, perhaps not so accurate. I also did a comparison between Dmitry's model against the actual tube curves from Audiomatica, no surprise, they match up very well, too bad, it only goes up to 0.6A.

An externally hosted image should be here but it was not working when we last tested it.

 

[audiodesign]

I have converted the model in LTSpice but it does not run good

.SUBCKT V6C33C-B 1 2 3
* A G C
*
* X1 1 2 3 TRIO3 {G=5.126m MU=2 B=18.79 C=.944 Vc=.22}
*X2 2 3 Igrid {ALPHA=1m BETA=5U}
X1 1 2 3 TRIO3
X2 2 3 Igrid
C1 1 2 10PF
C2 2 3 20PF
C3 1 3 7PF
*
.ENDS

.SUBCKT Igrid G1 C
+ PARAMS: ALPHA=1m BETA=5U
* Courant grille
B4 6 0 V=V(G1,C)>0 ?{ALPHA}*V(G1,C)^1.5 : {BETA}/-(V(G1,C)-.1)
B5 G1 C I=V(6,0)
.ENDS

.SUBCKT TRIO3 A G C
+ PARAMS: G=5.126m MU=2 B=18.79 C=.944 Vc=.22
B1 A C I={G}*(1+(V(G,C)/({B}-(V(G,C)/{C}))))*(V(G,C)+((V(A,C)+{Vc})/{MU}))^1.5
.ENDS

 

[audiodesign]

Upon further reading - the Excem model is the same as the Rydel model (with some adjustments to the parameters), so as you said, perhaps not so accurate. I also did a comparison between Dmitry's model against the actual tube curves from Audiomatica, no surprise, they match up very well, too bad, it only goes up to 0.6A.

An externally hosted image should be here but it was not working when we last tested it.

This is not bad, which model have you used ?

 

[tubetvr]

the Excem model is the same as the Rydel model

No not exactly, there is an additional term (VE in the formula C in the model) to introduce the kink see formula (3) in here http://www.excem.fr/download/usergui5.pdf

The problem is that the kink is not so smooth as predicted in reality the curves bends quite abrupt at a specific low current, you can see it best at around -100V grid voltage, what is needed is either a polynom with much higher order or a piecewise linear approximation for small steps of current.

 

[audiodesign]

I have converted the model in LTSpice but it does not run good

.SUBCKT V6C33C-B 1 2 3
* A G C
*
* X1 1 2 3 TRIO3 {G=5.126m MU=2 B=18.79 C=.944 Vc=.22}
*X2 2 3 Igrid {ALPHA=1m BETA=5U}
X1 1 2 3 TRIO3
X2 2 3 Igrid
C1 1 2 10PF
C2 2 3 20PF
C3 1 3 7PF
*
.ENDS

.SUBCKT Igrid G1 C
+ PARAMS: ALPHA=1m BETA=5U
* Courant grille
B4 6 0 V=V(G1,C)>0 ?{ALPHA}*V(G1,C)^1.5 : {BETA}/-(V(G1,C)-.1)
B5 G1 C I=V(6,0)
.ENDS

.SUBCKT TRIO3 A G C
+ PARAMS: G=5.126m MU=2 B=18.79 C=.944 Vc=.22
B1 A C I={G}*(1+(V(G,C)/({B}-(V(G,C)/{C}))))*(V(G,C)+((V(A,C)+{Vc})/{MU}))^1.5
.ENDS

Yes, this does not work on LTSpice, how to use it ?

 

[tubetvr]

Yes, this does not work on LTSpice, how to use it ?

I have never used LTSpice is it based on PSpice? I use something called Superspice that is a version based on Spice 3 and then it works.

 

[Osvaldo de Banfield]

What does a .22R resistor do at the cathode of a tube? This value and a wire are almost equal, I believe...

 

[Dimitris AR]

What does a .22R resistor do at the cathode of a tube? This value and a wire are almost equal, I believe...

This resistor I believe ( even if it is low value ) will reduce a little bit the high transconductance of the 6C33C and make it more linear , see degenerative ! .

 

[12E1]

What does a .22R resistor do at the cathode of a tube? This value and a wire are almost equal, I believe...

With the tubes in parallel with separate resistors, the small amount of degenerative feedback from the resistor helps to equalise the current sharing between the two valves, making the design more tolerant of slight mis-matches in their curves and / or biasing. (The same technique is also commonly used when transistors are used in a parallel arrangement.)