开云体育

  1. LTspice
  2. Topics

Interpreting Noise Simulation Results

 

Hello,
?
I don't have much experience using LTSPICE for noise simulations, and I'm having trouble understanding some of the results I'm getting. I'm investigating an amplifier circuit (file "Amplifier Noise" uploaded to files/Temp).
?
The amplifier is a cascode differential pair, signal frequency is 1MHz. Currently the circuit has no filtering besides the inherent roll-off at high frequency. I'm investigating whether adding AC coupling capacitors to the input to filter low-frequency noise can improve the performance. The inductance/resistance in series with each input represents the output impedance of the previous stage.
?
When I simulate the original circuit with no coupling capacitors, the result seems reasonable - relatively flat, input noise is a fraction of the output noise. However when I simulate the proposed circuit with coupling capacitors I don't understand the result.
?
The input noise balloons at low frequencies - the frequency range that the HPF of the capacitor is supposed to reduce, is now much greater. The output noise is lower in terms of total integrated noise, but still has an increased magnitude at low frequency compared to the original.
?
The decrease in total output noise is coming from a decrease in the high frequency noise, which makes sense because the inductance of the source impedance and the base pull-down resistor form an RL LPF. But the capacitor does not seem to be reducing LF noise, even though an AC sweep shows the expected HPF behavior.
?
Am I misinterpreting the results somehow? How is adding a high-pass filter at the input increasing noise at the low frequencies that it is supposed to be attenuating? What am I missing about noise simulations?
 

开云体育

At low frequencies, the capacitor's impedance is quite high. the noise current generated by the active stages circulates in this increased impedance, resulting in an increasing voltage.

Le 20/02/2025 à 20:17, manauo via groups.io a écrit?:

Hello,
?
I don't have much experience using LTSPICE for noise simulations, and I'm having trouble understanding some of the results I'm getting. I'm investigating an amplifier circuit (file "Amplifier Noise" uploaded to files/Temp).
?
The amplifier is a cascode differential pair, signal frequency is 1MHz. Currently the circuit has no filtering besides the inherent roll-off at high frequency. I'm investigating whether adding AC coupling capacitors to the input to filter low-frequency noise can improve the performance. The inductance/resistance in series with each input represents the output impedance of the previous stage.
?
When I simulate the original circuit with no coupling capacitors, the result seems reasonable - relatively flat, input noise is a fraction of the output noise. However when I simulate the proposed circuit with coupling capacitors I don't understand the result.
?
The input noise balloons at low frequencies - the frequency range that the HPF of the capacitor is supposed to reduce, is now much greater. The output noise is lower in terms of total integrated noise, but still has an increased magnitude at low frequency compared to the original.
?
The decrease in total output noise is coming from a decrease in the high frequency noise, which makes sense because the inductance of the source impedance and the base pull-down resistor form an RL LPF. But the capacitor does not seem to be reducing LF noise, even though an AC sweep shows the expected HPF behavior.
?
Am I misinterpreting the results somehow? How is adding a high-pass filter at the input increasing noise at the low frequencies that it is supposed to be attenuating? What am I missing about noise simulations?
 

开云体育

Yes, an RC high pass filter on the input will increase the noise. Although the capacitor doesn't itself add any noise, it causes the effective input noise voltage to increase, because:

Vn(tot) = √(En^2 + (In/2/pi/Cser)^2)

..where: En = I/P noise voltage density, and In = I/P noise current density (neglecting the source resistance noise, which often you can't do)

Remember also: "No attenuation before gain".

I should also mention (without seeing your schematic), that many of the devices in the LTspice standard libraries don't have realistic noise parameters, especially when it comes to 1/f noise.

--
Regards,
Tony


On 20/02/2025 20:17, manauo via groups.io wrote:

I don't have much experience using LTSPICE for noise simulations, and I'm having trouble understanding some of the results I'm getting. I'm investigating an amplifier circuit (file "Amplifier Noise" uploaded to files/Temp).
?
The amplifier is a cascode differential pair, signal frequency is 1MHz. Currently the circuit has no filtering besides the inherent roll-off at high frequency. I'm investigating whether adding AC coupling capacitors to the input to filter low-frequency noise can improve the performance. The inductance/resistance in series with each input represents the output impedance of the previous stage.
?
When I simulate the original circuit with no coupling capacitors, the result seems reasonable - relatively flat, input noise is a fraction of the output noise. However when I simulate the proposed circuit with coupling capacitors I don't understand the result.
?
The input noise balloons at low frequencies - the frequency range that the HPF of the capacitor is supposed to reduce, is now much greater. The output noise is lower in terms of total integrated noise, but still has an increased magnitude at low frequency compared to the original.
?
The decrease in total output noise is coming from a decrease in the high frequency noise, which makes sense because the inductance of the source impedance and the base pull-down resistor form an RL LPF. But the capacitor does not seem to be reducing LF noise, even though an AC sweep shows the expected HPF behavior.
?
Am I misinterpreting the results somehow? How is adding a high-pass filter at the input increasing noise at the low frequencies that it is supposed to be attenuating? What am I missing about noise simulations?
 

manauo,
?
Almost all the simulated noise in your circuit comes from the circuit itself.? If there was noise in the sources driving it (V1, V2, V5, V6), it is not present in the simulation.? Those voltage sources are noise-free.? Therefore, adding high-pass filter capacitors to reduce noise from the source, does not do you any good, unless the source itself had noise.? I think the only parts behind the HPF that generate any noise in this simulation, are the 0.6 ohm resistors, R9, R10, R21, and R22.? (Also the Rser of L1-L4.)? That's about the only benefit to be had from adding those capacitors.
?
On top of that, there is the increase in noise caused by making the source impedance (connected to the transistor bases) larger.? That is likely why you saw the noise density going up instead of down.
?
Not many people use SPICE's .NOISE analysis.? Consequently, many device models either lack the parameters needed to simulate it well, or the parameter values are badly wrong.? That was especially true of the JFET models that came with LTspice.? LTspice got those models from the company Linear Systems (not to be confused with Linear Technology Corp.), but someone at Linear Systems had inadvertently messed up their own SPICE models before giving them to LTC for inclusion in LTspice.? If I remember correctly, the mistake was 5 orders of magnitude (!) and applied to every one of their JFET models.? Fortunately your circuit has only BJTs, no JFETs, so it might be OK.? Should be OK for comparative simulations.
?
Andy
?
 

开云体育

Further to what I wrote below, I have uploaded a modified version of your schematic that uses a single amplifier circuit but switches the filter networks into or of out of circuit. This enables the noise with and without the HPFs to be displayed on the same graph, by stepping the Ctl parameter.

You should also know that the FMMT718 model you have added is unreasonably optimistic, as it exhibits no 1/f noise. In any event it is swamped by the increased noise from the HPFs, when these are in circuit.

You could drastically decrease the noise contribution of the HPFs by scaling the impedance down, while keeping the cut-off frequency the same. I've shown the effect of changing Rfilt from 100Ω to 10kΩ.

--
Regards,
Tony


On 20/02/2025 23:19, Tony Casey wrote:

Yes, an RC high pass filter on the input will increase the noise. Although the capacitor doesn't itself add any noise, it causes the effective input noise voltage to increase, because:

Vn(tot) = √(En^2 + (In/2/pi/Cser)^2)

..where: En = I/P noise voltage density, and In = I/P noise current density (neglecting the source resistance noise, which often you can't do)

Remember also: "No attenuation before gain".

I should also mention (without seeing your schematic), that many of the devices in the LTspice standard libraries don't have realistic noise parameters, especially when it comes to 1/f noise.

--
Regards,
Tony


On 20/02/2025 20:17, manauo via groups.io wrote:
I don't have much experience using LTSPICE for noise simulations, and I'm having trouble understanding some of the results I'm getting. I'm investigating an amplifier circuit (file "Amplifier Noise" uploaded to files/Temp).
?
The amplifier is a cascode differential pair, signal frequency is 1MHz. Currently the circuit has no filtering besides the inherent roll-off at high frequency. I'm investigating whether adding AC coupling capacitors to the input to filter low-frequency noise can improve the performance. The inductance/resistance in series with each input represents the output impedance of the previous stage.
?
When I simulate the original circuit with no coupling capacitors, the result seems reasonable - relatively flat, input noise is a fraction of the output noise. However when I simulate the proposed circuit with coupling capacitors I don't understand the result.
?
The input noise balloons at low frequencies - the frequency range that the HPF of the capacitor is supposed to reduce, is now much greater. The output noise is lower in terms of total integrated noise, but still has an increased magnitude at low frequency compared to the original.
?
The decrease in total output noise is coming from a decrease in the high frequency noise, which makes sense because the inductance of the source impedance and the base pull-down resistor form an RL LPF. But the capacitor does not seem to be reducing LF noise, even though an AC sweep shows the expected HPF behavior.
?
Am I misinterpreting the results somehow? How is adding a high-pass filter at the input increasing noise at the low frequencies that it is supposed to be attenuating? What am I missing about noise simulations?
 

I didn't consider the impact of the high impedance of the capacitor increasing the current noise at the input - thank you!
 

Thanks for the insight - I read that the voltage sources were treated as noiseless but when I set one of them as the input I lost sight of that. Appreciate the help!

1 - 7 of 7