Jfet Spice Data CTC 036 Interfet
Jfet Spice Data CTC 036 Interfet
Jfet Spice Data CTC 036 Interfet
DC Model
The DC model parameters (𝑉𝑇0 , 𝛽, and 𝜆) are the most crucial set of parameters needed to
generate a SPICE model of a JFET. These values are expressed in SPICE2 when performing ‘Id-Vds’ or ‘Id-
Vgs’ sweeps. The SPICE2 implementation of the JFET static model uses the Shichman-Hodges polynomial
models to depict the Cutoff, Ohmic, and Saturation regions of JFETs as separate piecewise equations,
shown in equation 1.
Figure 1, Using an ‘Id-Vgs’ sweep (N0132SL) to find a linear regression for obtaining VTO and
BETA.
Computing 𝜆 is a little different and requires 𝑉𝑇0 and 𝛽 to be calculated first. Once 𝑉𝑇0 and 𝛽 are
known, take an ‘Id-Vds’ sweep, remove all data points with abs(𝑉𝐷𝑆 ) less than abs(𝑉𝐺𝑆 (𝑂𝐹𝐹)). I usually
remove all points with abs(𝑉𝐷𝑆 ) below ~1.2 ∙ abs(𝑉𝐺𝑆 (𝑂𝐹𝐹)) to ensure all remnants from the “knee” of
the sweep are removed. Then perform Equation 4 to obtain 𝜆.
𝐼𝐷 (𝑉𝐷𝑆 , 𝑉𝐺𝑆 ) 1
𝜆 = mean ( 2
− )
𝛽𝑉𝐷𝑆 (𝑉𝐺𝑆 − 𝑉𝑇𝑂 ) 𝑉𝐷𝑆
𝑉𝐺𝑆 −𝑚
𝐶𝐺𝑆𝑖 (1 − ) 𝑉𝐺𝑆 ≤ (𝐹𝐶 ∙ 𝜑0 ) (𝑅𝑒𝑣𝑒𝑟𝑠𝑒 𝐶𝑎𝑝𝑎𝑐𝑖𝑡𝑖𝑣𝑒)
𝜑0
𝐶𝐺𝑆 =
𝐶𝐺𝑆𝑖 𝑚𝑉𝐺𝑆
1+𝑚
((1 − 𝐹𝐶 )(1 + 𝑚) − ) 𝑉𝐺𝑆 > (𝐹𝐶 ∙ 𝜑0 ) (𝐿𝑖𝑛𝑒𝑎𝑟 𝐶𝑎𝑝𝑎𝑐𝑖𝑡𝑖𝑣𝑒)
{(1 − 𝐹𝐶 ) 𝜑0
𝑉𝐺𝐷 −𝑚
𝐶𝐺𝐷𝑖 (1 − ) 𝑉𝐺𝐷 ≤ (𝐹𝐶 ∙ 𝜑0 ) (𝑅𝑒𝑣𝑒𝑟𝑠𝑒 𝐶𝑎𝑝𝑎𝑐𝑖𝑡𝑖𝑣𝑒)
𝜑0
𝐶𝐺𝐷 =
𝐶𝐺𝐷𝑖 𝑚𝑉𝐺𝐷
1+𝑚
((1 − 𝐹𝐶 )(1 + 𝑚) − ) 𝑉𝐺𝐷 > (𝐹𝐶 ∙ 𝜑0 ) (𝐿𝑖𝑛𝑒𝑎𝑟 𝐶𝑎𝑝𝑎𝑐𝑖𝑡𝑖𝑣𝑒)
{(1 − 𝐹𝐶 ) 𝜑0
𝑉𝐺𝑆 − 𝑉𝐷𝑆
𝜑0 = MEAN 2
𝐶𝑖
1−( )
( 𝐶𝑟𝑆𝑆 (𝑉𝐺𝑆 , 𝑉𝐷𝑆 ) )
Equation 9, Equation to solve for the Gate-Junction Potential.
Gate Leakage
JFET Gate Leakage is modeled in SPICE2 similarly to that of ideal PN-Junction diodes, which
additionally carries into how JFETs equivalent diodes are modeled in SPICE2. There is only one SPICE
parameter involved in this calculation, being the Gate Saturation Current Parameter 𝐼𝑆 (not to be
confused with the JFET Source Current). There is also the leakage conductance SPICE value 𝐺𝑀𝑖𝑛 that can
be modified separately, but unfortunately that impacts all models that utilize 𝐺𝑀𝑖𝑛 in that given
simulation, including other JFETs and diodes. The default value of 𝐺𝑀𝑖𝑛 in SPICE2 is 10−12 Siemens.
The leakage (and overall diode behavior) is modeled as two diodes, Gate to Drain and Gate to
Source. The equations for the currents traveling through the diodes is shown in Equations 8-9. When a
JFET is configured as a two-terminal diode, either of these equations can be used if the equation is
𝑘𝐵
doubled and 𝑉𝐺𝐷 or 𝑉𝐺𝑆 are treated as 𝑉𝐷 (the voltage applied to the diode). Note that =
𝑞
8.617333 ∙ 10−5 𝑒𝑉/°𝐾.
𝑘𝐵 𝑇
−𝐼𝑆 + 𝑉𝐺𝐷 𝐺𝑀𝑖𝑛 𝑉𝐺𝐷 ≤ −5 (𝐿𝑖𝑛𝑒𝑎𝑟 𝐿𝑒𝑎𝑘𝑎𝑔𝑒)
𝑞
𝐼𝐺𝐷 = 𝑘 𝑇 −1
𝑉𝐺𝐷 ( 𝐵 ) 𝑘𝐵 𝑇
𝐼𝑆 (𝑒 𝑞 − 1) + 𝑉𝐺𝐷 𝐺𝑀𝑖𝑛 𝑉𝐺𝐷 > −5 (𝐸𝑥𝑝𝑜𝑛𝑒𝑛𝑡𝑖𝑎𝑙 𝐿𝑒𝑎𝑘𝑎𝑔𝑒)
{ 𝑞
Equation 10, SPICE equation for the Gate to Drain Leakage Current.
𝑘𝐵 𝑇
−𝐼𝑆 + 𝑉𝐺𝑆 𝐺𝑀𝑖𝑛 𝑉𝐺𝑆 ≤ −5 (𝐿𝑖𝑛𝑒𝑎𝑟 𝐿𝑒𝑎𝑘𝑎𝑔𝑒)
𝑞
𝐼𝐺𝑆 = 𝑘 𝑇 −1
𝑉𝐺𝑆 ( 𝐵 ) 𝑘𝐵 𝑇
𝐼𝑆 (𝑒 𝑞 − 1) + 𝑉𝐺𝑆 𝐺𝑀𝑖𝑛 𝑉𝐺𝑆 > −5 (𝐸𝑥𝑝𝑜𝑛𝑒𝑛𝑡𝑖𝑎𝑙 𝐿𝑒𝑎𝑘𝑎𝑔𝑒)
{ 𝑞
Equation 11, SPICE equation for the Gate to Source Leakage Current.
Since the Forward-Bias Voltage (𝑉𝐹𝐵 ) of the diodes isn’t directly controllable, a compromise may
have to be made between choosing an accurate gate leakage current and accurate 𝑉𝐹𝐵 . Additionally, it
should be noted that the diode breakdown voltage and current is not included in the JFET diode model.
Figure 14 shows a diode characteristic of an example JFET using the SPICE2 equations using an N-channel
J109.
Figure 3, An ‘Ig-Vgs’ sweep (N0450SL) showing the JFET diode behavior in SPICE2.
The 𝐼𝑆 parameter can be computed alongside 𝐺𝑀𝑖𝑛 (default 𝐺𝑀𝑖𝑛 = 10−12 Siemens) or be
computed independently. By shorting the Source and Drain pins of the FET and reverse biased the FET, a
reverse-bias diode characteristic will be formed, and can be used in the following equations for
computation:
𝐼𝑆 = 𝑉𝐷 𝐺𝑀𝑖𝑛 − 𝐼𝐺 (𝑉𝐷 )
Equation 12, Equation for computing Gate Saturation Current Parameter.
For both the 𝐼𝑆 and 𝐺𝑀𝑖𝑛 parameters, a full reverse-bias sweep will be needed, and to perform a linear
regression to obtain a slope (𝑚𝑔 ) and an intercept (𝑏𝑔 ):
𝐼𝑆 = −𝑏𝑔 𝐺𝑀𝑖𝑛 = 𝑚𝑔
Equation 13, Equation for computing Gate Saturation Current Parameter and the Leakage
Conductance.
Parasitic Resistances
All real JFETs exhibit parasitic resistances to varying extents, because of bond wires,
semiconductor resistances, lead resistances, etc.. Working with parasitic resistances is tricky, because the
parasitic resistances are not possible to model with a single equation and instead need to be simulated.
Figure 4, A JFET model showing parasitic resistances externally. Schematic produced using
LTSpice.
Even though the parasitic resistances are typically less than 10 Ohms, the impact of these
resistances can be significant, especially on parts with higher 𝐼𝐷𝑆𝑆 . Figures 16-18 show the impact of
parasitic resistances on a high-current JFET, note the impact of 𝑟𝐷 on the observed knee position and the
impact of 𝑟𝑆 on 𝐼𝐷𝑆𝑆 . There may be a way of determining the parasitic resistances if 𝑟𝐷 ≈ 𝑟𝑆 that involves
performing a few ‘Vgs-Id’ sweep that includes a slightly positive 𝑉𝐺𝑆 to counteract the effect of 𝑟𝑆 .
It should be noted that the parasitic resistances impact the large signal, small signal, and noise
behavior of a JFET in SPICE2. As a result, modeling the parasitic resistances with precision and accuracy is
a critical task.
Figure 5, A comparative simulation of an IF3601 with no parasitic resistance (green) and with
a 1 Ohm drain resistance (red). Simulated using LTSpice.
Figure 6, A comparative simulation of an IF3601 with no parasitic resistance (green) and with
a 1 Ohm source resistance (red). Simulated using LTSpice.
Figure 7, A comparative simulation of an IF3601 with no parasitic resistance (green) and with
1 Ohm drain and source resistances (red). Simulated using LTSpice.
Figure 8, An equivalent small-signal model for depicting SPICE2 JFET noise current sources.
The names of each of the current sources shown depict their noise component, excluding the
current source labeled “Channel,” which is a non-noise current source for depicting the small-signal
current traveling through the JFET. Then the question becomes, what are Flicker, Shot, and Johnson
noise? It is worth noting that 𝑘𝐵 = 1.380649 ∙ 10−23 𝐽/°𝐾 in this section.
Flicker noise is a noise contribution with the cause not fully understood, acting on the device
channel. It is the only noise contribution with SPICE parameters exclusively for this function, 𝐴𝐹 and 𝐾𝐹 .
This noise source is impacted by the drain current and decays inversely with frequency, because of this, it
is most prominent in lower frequencies [2].
𝐾𝐹 𝐼𝐷 𝐴𝐹
𝑁𝐹𝑙𝑖𝑐𝑘𝑒𝑟 (𝐴2 ) = ∆𝑓
𝑓
Equation 14, SPICE equation for the Channel Flicker Noise contribution.
Shot noise (also known as Poisson) noise is another contribution induced on the device channel.
This component is caused by the discrete and chaotic behavior of moving electrical charges. This noise
source is impacted by the junction temperature and the transconductance of the device at any given
time. This component is present and uniform across all frequencies [2].
4𝑘𝐵 𝑇 4𝑘𝐵 𝑇
𝑁𝐷𝑟𝑎𝑖𝑛−𝑇ℎ𝑒𝑟𝑚𝑎𝑙 (𝐴2 ) = ∆𝑓 𝑁𝑆𝑜𝑢𝑟𝑐𝑒−𝑇ℎ𝑒𝑟𝑚𝑎𝑙 (𝐴2 ) = ∆𝑓
𝑟𝐷 𝑟𝑆
Equation 16, SPICE equations for the Thermal Noise contributions of the Parasitic Resistances.
All the noise current sources occur in parallel to some form of a resistance, as a result, they can
be expressed as Thevenin-Norton Equivalent voltage sources. It is critical to note that the Shot and
Flicker noise components share a parallel resistance, and as a result can be added together.
So, once the values for all the noise components at any given temperature, drain current, and
transconductance have been determined, how can that be converted to Input Referred Voltage Noise?
First, here’s a sample circuit for computing the noise. Output-Referred Noise is noise occurring at the
drain, and Input-Referred Noise is the noise occurring at the gate.
Figure 9, A simple circuit for measuring Voltage or Current Noise in LTSpice. Schematic
produced using LTSpice.
To compute the noise at different nodes, we need to convert all the current noise sources into the
Output-Referred Current Noise:
Equation 17, Approximate equation for the SPICE Output-Referred Current Noise.
After doing that, we can then calculate the Output or Input-Referred Voltage Noise:
𝑉 𝑉 𝑁𝑂𝑅−𝑉
𝑁𝑂𝑅−𝑉 ( ) ≈ 𝑁𝑂𝑅−𝐼 𝑅𝐷𝐸 → 𝑁𝐼𝑅−𝑉 ( )≈
√𝐻𝑧 √𝐻𝑧 𝑔𝑚 𝑅𝐷𝐸
Equation 18, Approximate equations for the SPICE Output and Input-Referred Voltage Noise.
To calculate the SPICE parameters needed for accurate noise behavior, you first need to take the
measured noise (input-referred) and convert it into output-referred current noise.
Equation 19, Approximate equations for the SPICE Output and Input-Referred Voltage Noise.
Equation 20, Approximate equations for the SPICE Output and Input-Referred Voltage Noise.
Then compute the following values (𝑥𝑉 and 𝑦𝑉 ) and perform a linear regression on them,
obtaining the slope (𝑚) and y-intercept (𝑏).
𝑥𝑉 = 𝑙𝑛(𝐼𝐷 ) 𝑦𝑉 = 𝑙𝑛(𝑁𝐹𝑙𝑖𝑐𝑘𝑒𝑟−𝐶𝑎𝑙𝑐 (𝐼𝐷 , 𝑓) ∙ 𝑓)
𝑎𝐹 = median(𝑚) 𝑘𝐹 = median(𝑒 𝑏 )
3E-09
Voltage Noise (nV/rHz)
5E-10
0
10 100 1000 10000
Frequency (Hz)
Figure 10, An Input-Referred Voltage Noise Sweep with overlayed SPICE model.
Citations
[1] G. Massobrio and P. Antognetti, Semiconductor Device Modeling with Spice. McGraw-
Hill, 2009.
[2] R. Keim, What Is Electrical Noise and Where Does It Come From?. All About Circuits,
2018.