# Why does LTspice resistor noise disappear when setting value with .param?

When performing a noise analysis in LTspice of a 1kΩ resistor, a noise level of 4nV/rtHz is simulated.

However, if this resistor value is set to 1kΩ using .param, its noise decreases to 129fV/rtHz. Changing this resistor value does change the noise linearly. What causes this noise decrease, and what can I do to prevent it?

The schematic is shown below:

• At 20 degC, 1k gives 4nV/rtHz but 1 ohm gives 0.127 nV/rtHz. Maybe the "k" is upsetting things. Try 1000 in the .param thing. – Andy aka Jan 3 '20 at 14:29

You should use {Rg} as value for R1, so without the R=.

Using the plain value 1u or R={1k} also gives 129 fV/Hz½.

I've no idea what makes the evaluation of R={1k} or R={Rg} to become 1uΩ

• In that case, how can I use arithmetic in the value field, such as: i.imgur.com/bvczWO8.png This raises the error: "Unknown parameter '-' ". – Elmardus Jan 3 '20 at 14:53
• To answer my own question: mathematical operations should also be within the curly braces, not outside them. So, to set the resistor for my circuit the value should be set to: {350-Rg} – Elmardus Jan 3 '20 at 15:09
• R= replaces the primitive resistor with a behavioural resistor, which no longer has a builtin noise source. And, as you perhaps know, if you don't need noise in an R/L/C, add noiseless next to the value. The small noise figure is just floating point residual. @Elmardus Elements should have values, and curly braces (or single quotes) perform evaluation such that the resulting value is passed to the parser and further. Behavioural elements (V=, I=, R=) don't need curly braces because the expression after the equal is automatically understood as an expression to be evaluated. – a concerned citizen Jan 4 '20 at 20:12

In LTSPICE XVII(64) this seems to work... Noise output is about 0.9nV/rt(Hz) as expected:
Perhaps the 129fV/rt(Hz) is a result of an added Thevenin resistance added to the voltage source??