Hi all,
As I am doing my experiments with valves, many times I found my amplifiers oscillating.
To get a better insight of what is contributing to this, I have decided to simulate the circuit in LTSPICE, however the results I get make no sense at all.
Just to show an example, I designed a basic 2 stage 12AY7 pre-amp with the usual values for the resistors, capacitors, etc.
To do the analysis, I used the normal approach:
1 - Connected the amp in unitary feedback
2 - Broke the feedback puting a large resistor and capacitor, followed by a DC voltage source with DC=0 and AC=1
3 - AC analysis with 50 points per decade from 100Hz to 10MHz
Simulating the circuit gives me non-sense results. The poles should contribute with a 20dB/dec roll-off and 90deg/dec, which doesn't happen. It can be seen that the magnitude curve never crosses 0dB.
What am I doing wrong?
Do the models of the valves behave correctly for small signal?
Kind regards,
Pedro
As I am doing my experiments with valves, many times I found my amplifiers oscillating.
To get a better insight of what is contributing to this, I have decided to simulate the circuit in LTSPICE, however the results I get make no sense at all.
Just to show an example, I designed a basic 2 stage 12AY7 pre-amp with the usual values for the resistors, capacitors, etc.
To do the analysis, I used the normal approach:
1 - Connected the amp in unitary feedback
2 - Broke the feedback puting a large resistor and capacitor, followed by a DC voltage source with DC=0 and AC=1
3 - AC analysis with 50 points per decade from 100Hz to 10MHz
Simulating the circuit gives me non-sense results. The poles should contribute with a 20dB/dec roll-off and 90deg/dec, which doesn't happen. It can be seen that the magnitude curve never crosses 0dB.
What am I doing wrong?
Do the models of the valves behave correctly for small signal?
Kind regards,
Pedro
Last edited:
I made an error when plotting.
I tried converting to logarithmic, but the plot was already logarithmic. So, doing 20*log10(Vin) is wrong !
I tried converting to logarithmic, but the plot was already logarithmic. So, doing 20*log10(Vin) is wrong !
Simple enough circuit, name all the nodes (makes results easier to follow). Then look at all the DC bias points. Your answer will be there. Steveu has pointed out the likely problem.
I think you are mixing two different strategies for measuring OLG.
1. The first simple strategy drives the input with no feedback and compares the output to the input. For op-amps, DC feedback is required to set the operating point, so a huge capacitor is used to remove the AC feedback. But your tubes operating point is set by the cathode resistors, and no DC feedback is needed. The grid DC voltage needs to be ~zero. Note as well that your feedback is positive. In a real circuit, the output would probably clip so this is not the preferred approach, although Simulation can ignore such limitations.
2. The preferred approach is to put the stimulus in the feedback loop. Then, you measure the amount that the output cancels the stimulus. There is no huge capacitor involved but the stimulus is floating. Suppose the OLG is x99 and the stimulus is 1V, then the output is 0.99V which cancels all but 0.01V of the stimulus, 0.01+0.99=1; 0.01x99=0.99.
3. If you really want to know about OLG simulation, you can read up on the TIAN probe, but it is complicated, and results are the same as a well-placed simple "probe". LTspice has some examples in the educational folder.
1. The first simple strategy drives the input with no feedback and compares the output to the input. For op-amps, DC feedback is required to set the operating point, so a huge capacitor is used to remove the AC feedback. But your tubes operating point is set by the cathode resistors, and no DC feedback is needed. The grid DC voltage needs to be ~zero. Note as well that your feedback is positive. In a real circuit, the output would probably clip so this is not the preferred approach, although Simulation can ignore such limitations.
2. The preferred approach is to put the stimulus in the feedback loop. Then, you measure the amount that the output cancels the stimulus. There is no huge capacitor involved but the stimulus is floating. Suppose the OLG is x99 and the stimulus is 1V, then the output is 0.99V which cancels all but 0.01V of the stimulus, 0.01+0.99=1; 0.01x99=0.99.
3. If you really want to know about OLG simulation, you can read up on the TIAN probe, but it is complicated, and results are the same as a well-placed simple "probe". LTspice has some examples in the educational folder.
Thanks both.
Yes, I was applying the same method as for the op-amps and, in this case, makes no sense.
Next try. The simplest one. I just applied the AC source to the grid of the first triode and measured at the output.
The cap I put at the output is floating but it is just to remove the dc value.
Yes, I was applying the same method as for the op-amps and, in this case, makes no sense.
Next try. The simplest one. I just applied the AC source to the grid of the first triode and measured at the output.
The cap I put at the output is floating but it is just to remove the dc value.