Checking spice model plate curve accuracy

[aletheian]

I have cobbled many spice models together to work in my Circuitmaker simulator, but I can't seem to generate a set of plate curves within the simulation to check the accuracy of my math.

i have tried setting up a triode model.. say a 12ax7... and sweeping the DC plate voltge parameter from 0-300v in incriments of 10v, while sweeping a negative grid voltage from 1-3v in incriments of .5v, but i can't seem to find a way in the program of generating a graph of the plate current against the plate voltage to actually SEE the curves.

I am running the student version, so it is possible that the parameters I need are not available, but it is more likely that i have just set up the simulation improperly.

Can anyone help out?

[aletheian]

I figured it out. I had it right, I was just zoomed in too far on the scope... DOH!

[aletheian]

i don't know if I am happy that i figured out my mistake or not... because now i have hard evidence that the models that i have been using are cr4ppy! I laid the generated curves over the data sheets and Sophia traces, and most of them don't match up. Some are startlingly close though... well, time for some good 'ol math!

[Giaime]

I never found a trusty ECC83/12AX7 model.

But I have extremely precise models of ECC88, 300B, 6SN7, etc etc...

all can be found here: http://www.geocities.com/koren_model/

See also here: Koren's spice models

[sawreyrw]

Are these models any good?

http://www.duncanamps.com/spicevalves.html

[Robert McLean]

I laid the generated curves over the data sheets and Sophia traces, and most of them don't match up.

I dont know exactly what SPICE models you are using, and I fully agree that some models are better than others, but to put it into perspective, what you may want to do is measure the plate characteristics of some actual 12AX7, for example, and compare them to the Sophia traces, or whatever other data you have. I think you may find many of them wont match up either. My point being that I see as much variation between actual tubes as between models.

[aletheian]

Originally posted by Robert McLean


I dont know exactly what SPICE models you are using, and I fully agree that some models are better than others, but to put it into perspective, what you may want to do is measure the plate characteristics of some actual 12AX7, for example, and compare them to the Sophia traces, or whatever other data you have. I think you may find many of them wont match up either. My point being that I see as much variation between actual tubes as between models.

Very true. I am trying to make some that are "close enough". I'll probably end up taking some sophia traces and data sheets and averaging the curves together and tracing them out.

If I can get a few of my staple tubes (12ax7, 6sn7, el84, el34, 6au6, ef86, 6922...etc) "close enough" to save a little breadboarding legwork, then I'll be happy.

Here is a confusing little jpg of the curves laid over a 12ax7 data sheet and sophia trace:

[aletheian]

Originally posted by Giaime
I never found a trusty ECC83/12AX7 model.

But I have extremely precise models of ECC88, 300B, 6SN7, etc etc...

all can be found here: http://www.geocities.com/koren_model/

See also here: Koren's spice models

I have not been able to hack Koren's models into a form that I can use. I can alter 3f4 and PSpice models to function by rearranging the syntax of the equasions, but the Koren models, unfortunately, won't work. If anyone has altered them into the eqiuvalent 3F4 form, then I'd be happy to give them a whirl.

[aletheian]

Originally posted by sawreyrw
Are these models any good?

http://www.duncanamps.com/spicevalves.html

i have used altered versions of the "generic triode model" as the basis for a lot of experiments. it is not too far off if you tweak it a bit and add a few more equasions. Actually, the triodes that I usually use are based on the Duncan anode model.

[Robert McLean]

I have not been able to hack Koren's models into a form that I can use. I can alter 3f4 and PSpice models to function by rearranging the syntax of the equasions, but the Koren models, unfortunately, won't work. If anyone has altered them into the eqiuvalent 3F4 form, then I'd be happy to give them a whirl.

If you search this forum for "Curve Captor" you will find a very useful program for generating triode spice models. It will output 3F4, Orcad, LTSpice, and Circuit Maker flavours of spice.

It also does 4 versions of Korens model, as well as others

Here is what it spits out for a 6SN7 in Circuit Maker mode, just as an example, see if this works for you :

* 6SN7 CircuitMaker model
.subckt 6SN7 P G K
Bp P K I=(0.0231376265/1.0e3)*uramp(V(P,K)*ln(1.0+exp((7.300912403)+(7.30 0912403)*(21.38514806)*V(G,K)/V(P,K)))/(7.300912403))^(1.351490772)
.ends 6SN7