For resistors and semiconductors in Tina, it is not necessary to use an external noise source:
"TINA takes into account the noise sources inherent in resistors and semiconductors (transistors, diodes). Resistors are modeled as an ideal noiseless resistor in parallel with an equivalent thermal noise current. This equivalent noise is modeled by white noise, i.e., noise containing all frequency components with the same amplitude density. The contributions of all the noise generators in the circuit are calculated to determine each one’s magnitude at the output. The 'output noise' is formed from the square root of the sum of the squares (rms value) of these individual contributions."
I assume the other spice simulation environments are similar. I do not know about vacuum tube models.
"TINA takes into account the noise sources inherent in resistors and semiconductors (transistors, diodes). Resistors are modeled as an ideal noiseless resistor in parallel with an equivalent thermal noise current. This equivalent noise is modeled by white noise, i.e., noise containing all frequency components with the same amplitude density. The contributions of all the noise generators in the circuit are calculated to determine each one’s magnitude at the output. The 'output noise' is formed from the square root of the sum of the squares (rms value) of these individual contributions."
I assume the other spice simulation environments are similar. I do not know about vacuum tube models.
yes, LTSpice calculates noise for resistors, capacitors (if they have ESR set), may be semiconductors. But for tubes it does not. I suspect that tube models do not include some "noise interface" that presents its noise to SPICE. It would be nice to add that "noise interface" and that should add noise as R calculated by Merlin's formulae with Sigma and K specific to a tube. I guess that this R should be exported as a Rp. Then SPICE should pick it up a standard way to bring into overall noise picture.
I can't do that in Tina because it won't allow more than one source. Oh well. Thanks anyway.
Do you need more than one? You probably need to examine only one source at a time anyway... (I use Tina too)
I guess not. It can become labor intensive to do one tube at a time, depending on how many you have. I'm still getting unrealistic results using this technique. With the input shorted by a 200 ohm resistor and one 2nA current source in parallel with one tube in the circuit I'm analysing, I'm getting 140 dB SNR, which isn't possible... Maximum should be 129 or so. I guess I don't understand how to use Tina very well.
Attachments
The current sources would need to be set to produce an actual noise signal, rather than a sine wave, for the SNR feature to understand them. I don't think you can do that.
If you want to use the SNR feature then you're probably better off inserting a resistor in series with each grid, to serve as an equivalent noise resistance for each tube. Something between 500 ohms and 3k ohms is realistic, depending on the tube you will use. This won't model the noise spectrum correctly (no pink noise), but if you just want SNR it's OK.
If you want to use the SNR feature then you're probably better off inserting a resistor in series with each grid, to serve as an equivalent noise resistance for each tube. Something between 500 ohms and 3k ohms is realistic, depending on the tube you will use. This won't model the noise spectrum correctly (no pink noise), but if you just want SNR it's OK.
It is also possible in Tina to add V or I source that is loaded from noise wav file.
You can find white, pink and other color noise files here High Definition Audio Test Files
You can do Transient Analysis, however Noise Analysis does not accept multiple sources.
You can find white, pink and other color noise files here High Definition Audio Test Files
You can do Transient Analysis, however Noise Analysis does not accept multiple sources.
From various sources I've found that flicker noise is already embedded into SPICE2 diode model - take a look at parameters KF (flicker noise coefficient) and AF (flicker noise exponent).
Many tube models use diode model. So may be it is possible to add these parameters to there?
I tried with 6c45pe in LTSpice and TINA-TI, they change noise with changing KF and AF.
Now it is only required to translate Merlin's K and Sigma into KF and AF and we will have noise analysis of tubed circuits.
Unfortunately I do not have enough knowledge how to do it.
Merlin could you take a look?
example of 6S45P with noise
Many tube models use diode model. So may be it is possible to add these parameters to there?
I tried with 6c45pe in LTSpice and TINA-TI, they change noise with changing KF and AF.
Now it is only required to translate Merlin's K and Sigma into KF and AF and we will have noise analysis of tubed circuits.
Unfortunately I do not have enough knowledge how to do it.
Merlin could you take a look?
example of 6S45P with noise
Last edited:
I am not sure. According to diode model it has RS (serial resistance) and the tube sets it to 1Ohm. So SPICE should automatically calculate thermal noise from this RS.
However there is another parameter TNOM (parameter measurement temperature) which is 27C by default. And the tube model does not specifies it.
May be it should be set to a real T inside tube? 1200C?
According to http://dalmura.com.au/projects/A study of noise in vacuum tubes and attached circuits.pdf
thermal noise will depend on Rp and T.
But in the tube model Rp is set to 1G. Huge. I am not sure if SPICE picks up that Rp.
I wish I knew more about this spice macro language and other things. Unfortunately, computer languages (and language in general) are not my forte.
I do know that with the tube models I have, which are from this message board and other popular places, the software becomes confused if there are no resistors in the circuit when I perform a noise analysis. In other words, if there's just a tube and some other things but no resistors, then I get really strange answers for the noise analysis. That leads me to believe that the spice models I have do not incorporate even thermal noise.
It would be a great service to the rest of us if you or someone else came up with something that solved this problem.
I do know that with the tube models I have, which are from this message board and other popular places, the software becomes confused if there are no resistors in the circuit when I perform a noise analysis. In other words, if there's just a tube and some other things but no resistors, then I get really strange answers for the noise analysis. That leads me to believe that the spice models I have do not incorporate even thermal noise.
It would be a great service to the rest of us if you or someone else came up with something that solved this problem.
Rp is not 1G, it is usually expressed as node "7 0" in the code, but it is merely used to make SPICE behave and has nothing to do with the thermal model. As Merlin pointed out, the external parts, e.g., Rg and/or the source resistance dominate tube voltage amplifier stage's noise, not the tube itself, so that's the first line of attack, after all, you can not change the tubes' parameter nor can you design or make new tubes now days like you can with SS devices.
well, but in the model it is still 1G
re1 7 0 1g
I am confused.
Also in the article http://dalmura.com.au/projects/A study of noise in vacuum tubes and attached circuits.pdf
author defines thermal noise based on Tk (cathode temperature), Rp and R load.
In tube model R load is not available, so thermal noise cannot be calculated right here.
re1 7 0 1g
I am confused.
Also in the article http://dalmura.com.au/projects/A study of noise in vacuum tubes and attached circuits.pdf
author defines thermal noise based on Tk (cathode temperature), Rp and R load.
In tube model R load is not available, so thermal noise cannot be calculated right here.
"7 0" means from node 7 to ground, it does not represent Rp which is the internal plate resistance. RL is the load on the output node, again, not a part of the tube itself. Unless you make them youself, thermal noise is usually not included in the tube models - there are of course some models that have thermal parameters built in, IIRC, such as the ones from Excem.
Last edited:
Sorry, I'm confused. Previously you said:
What is Rp in the model?
now you are saying:
What is Rp in the model?
I am not getting this, why then I hear hiss even with input shorted and without Rg? Why tubes are noisier than FETs? And why some tubes are noisier than others? I still believe that tubes not only amplify external noises, but also produce own noises in significant amount. These noises we are trying to fit into a model.
Returning to the article. There is Ratio of external resistance (RL) and internal tube resistance (Rp) in formula 2 and 3.
To me this ratio means a distribution of current between RL and Rp.
As soon as we know Ip inside tube model, can RL/Rp be somehow replaced by Ip? After all thermal noise depends on a current though R. Why then we cannot use Rp and Ip directly?
You are perhaps taking things out of context - of course, the tubes have noise, and some more than others, and usually they tend to be noisier than FETs, but still the relative noise level would be higher if you have Rg and/or Rs attached.
If Rp is not the internal plate resistance, then it must mean the resistor attached to the plate, i.e., an external resistor. Sorry but I did not read the paper from Tim's site...
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.