SPICE Tube Models — Beyond the Basics
Understanding Koren, Rydel, and Ayumi models for accurate tube simulation
SPICE was designed for semiconductors. Simulating vacuum tubes requires subcircuit models that replicate the tube's I-V characteristics using controlled sources. The three dominant model families each take a different approach to accuracy vs. complexity.
Koren Model Explorer
See how Koren parameters affect the predicted plate current
Koren Model IV Curves — 12AX7
SPICE Model Availability by Tube
Which models exist for your favorite tubes
| Tube | Koren | Rydel | Ayumi | Notes |
|---|---|---|---|---|
| 12AX7 | Yes | Yes | Yes | Most-modeled tube; all models accurate |
| 12AT7 | Yes | Yes | Yes | Good agreement across models |
| 12AU7 | Yes | Yes | Yes | Low-mu; Koren works well |
| 6SN7 | Yes | Some | Yes | Dual triode; match both sections |
| EL34 | Limited | Yes | Yes | Screen current model critical for pentode |
| KT88 | Limited | Yes | Yes | Ultralinear tap modeling adds complexity |
| 300B | Yes | Yes | Yes | DHT — filament as cathode, simple model |
| 6L6GC | Limited | Yes | Yes | Kink region needs special handling |
Setting Up Tube Simulations
1. Choosing a Simulator
LTspice (free, Windows/Mac) is the most popular for tube circuits. It handles nonlinear subcircuits well and converges reliably. SPICE OPUS, Ngspice, and TINA-TI are alternatives. For pentodes, ensure your simulator supports multi-terminal subcircuits with the .SUBCKT syntax.
2. Convergence Tips
Tube circuits often fail to converge in DC operating point analysis. Solutions: (1) Add .OPTIONS RELTOL=0.003 ITL1=500 to relax tolerances, (2) Use .IC statements to set initial node voltages near the expected operating point, (3) Add 1GΩ resistors from each node to ground to prevent floating nodes, (4) For transformer-coupled circuits, use K-statement coupling with realistic leakage inductance.
3. Validating Your Model
Always validate against published tube data: (1) Plot Ip vs Vp at several Vg values and compare to datasheet curves, (2) Check gm at the operating point against the datasheet value, (3) Verify mu by measuring ΔVp/ΔVg at constant Ip, (4) For pentodes, verify screen current is realistic (typically 20-40% of plate current). A model that matches at one operating point may diverge badly at extremes.
4. What SPICE Cannot Do
SPICE models don't capture: microphonics (mechanical vibration coupling), aging (emission decline over time), warm-up transients (cathode temperature rise), or grid emission at very high temperatures. For guitar amplifiers, the interaction between speaker impedance curves and output transformer characteristics is critical but hard to model accurately — always build and test.
Testez vos connaissances
What is the primary advantage of the Koren tube model over table-based models like Rydel?