5
\$\begingroup\$

I am attempting to simulate a long tail pair phase splitter using two 12AX7 triodes in LTSpice.

I tried multiple subcircuit models for the triode, all yielding the same results. I want to know if the circuit is wrong, or if the simulation is wrong.

The distortion of the signal is intentional, this is from a guitar amp with high distortion. I want to build this circuit, but as its from the internet and quite expensive, wanted to make sure it works and understand it by simulation.

Here is the relevant part from the schematic:

enter image description here

VC Approaches 250V, but on the transient reaches only about 120, still the problem is obvious. VA Approaches 310V VB Approaches 300V The phase splitter is supposed to drive a 6bq5 / EL84 push-pull pair.

And here is the transient analysis result.

enter image description here

What am I doing wrong either in the simulation or in the circuit?

\$\endgroup\$
4
  • \$\begingroup\$ I'd suggest removing global negative feedback from your early simulations. \$\endgroup\$
    – greg
    Jul 19 '15 at 15:21
  • \$\begingroup\$ @greg I have tried that, but it didn't affect the result. Couldn't it be, that the simulation is incorrect?(Perhaps because the triode subcircuit is just an approximation?) \$\endgroup\$
    – akaltar
    Jul 19 '15 at 19:11
  • \$\begingroup\$ I never had problems with simulating triodes long tail pairs using LTSpice. \$\endgroup\$
    – greg
    Jul 23 '15 at 17:31
  • \$\begingroup\$ @greg What subcircuit were you using? I tried both Duncan's and Norman Koren's (If I remember their names correctly) \$\endgroup\$
    – akaltar
    Jul 23 '15 at 21:20
2
\$\begingroup\$

What is the "Time to Start Saving Data"? It is very possible that the circuit will not work as expected for a few seconds because the simulation is also simulating thermal tube stuff that takes a while to stabilize. Maybe 10's of seconds. So your simulation should be like from 10s to 10.1s for 100ms.

\$\endgroup\$
4
  • 1
    \$\begingroup\$ The vacuum tube subcircuit I use doesn't take into account the heater, but I still tried your recommendation, and it seems you are right! After 10 seconds, both power valves are in use, even tough they are quite unbalanced, but that might be the reason this amp is supposed to have a unique sound. \$\endgroup\$
    – akaltar
    Aug 24 '15 at 14:21
  • 1
    \$\begingroup\$ I wouldn't rely too much on these simulations wrt how the circuit might sound. Most modern circuits use a lot of feedback to compensate for the highly nonlinear behavior of transistors. That fb makes circuits fairly predicable. But tube circuits usually don't use fb at all b/c they have limited gain and because tubes are actually significantly more linear than transistors so fb is less necessary (1.5 exponent vs 2 for transistor). This is a big clue as to why tube circuits sound different - if you overdrive a circuit with no fb, you get fewer high order harmonics. \$\endgroup\$
    – squarewav
    Aug 24 '15 at 23:08
  • \$\begingroup\$ I just wanted to test if the circuit is correct. Its supposedly of a famous model, so I can hear how it should sound. (Minus specific tubes, specific components etc.) But I understand your point. \$\endgroup\$
    – akaltar
    Aug 24 '15 at 23:29
  • 1
    \$\begingroup\$ Note that you can send real audio data into the input, and save the output to a file in LTspice. See this link for starters. \$\endgroup\$
    – rdtsc
    Mar 14 '16 at 18:17
1
\$\begingroup\$

I suspect C26 is in the wrong place. Where it is set in the circuit, if makes for a long RC constant for the bias of the right triode - in fact, I'm quite sure that is the time constant you are seeing in the plot. Also, if there is no C in the cathodes of the triodes, then the gain will only be 2.

Shouldn't C26 be in parallel with R54?

\$\endgroup\$
2
  • \$\begingroup\$ All the information I had about this kind of phase splitter is that the second grid should be AC grounded with a capacitor. Still I tried your suggestion and the result is that it works kind of! Just one of the tubes has a horrific clipping and they are unbalanced. Can post image if interested. Also as you can see from the naming, its the first tube that is not doing its job. \$\endgroup\$
    – akaltar
    Jul 18 '15 at 22:30
  • \$\begingroup\$ C26 is in the right place. It is not there to stabilize the bias. It's there to AC couple the fractional feedback signal (roughly devided by R65 and R56) into the 2ed input of this differential stage. \$\endgroup\$
    – Batuu
    Mar 14 '16 at 8:19
1
\$\begingroup\$

After actually building this thing and later coming back to make changes I realized I wasn't doing anything wrong at first either.

The simulation runs fine on LTspice's Alternate solver. I assume its a more stable and precise(albeit sometimes slower) solver. It gives the expected symmetrical curves (with some minor balance mismatch between the tubes).

Having a longer simulation with the normal solver also had better results than before, but it's still way off considering real life performance and the alternate solver.

TL;DR: Using the alternate solver seems to provide more precise information, and gave the expected results.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.