You are quite right. The static plate resistance is what counts at dc although its the dynamic one that is responsible for bias point stability.
Cheers
Ian
Yes, there are a number of minor errors. You will also have noticed that it is incomplete.
Cheers
Ian
Gents,
I'm still trying to get my head around the plate-to-grid feedback, and I went over Aiken's derivation using the 12AX7 as an example, the results matched pretty up well with the SPICE simulations. However, when using output pentode plate-to-grid feedback, the results no longer match, perhaps I am misapplying the equations or doing something stupid... Could someone please take a look and tell me what I did wrong... All I did was to look up the mu and rp for the pentode (6L6) and plug those figures into Aiken's equations, then compare them to the results from the SPICE simulation as shown below:
The operating parameters are:
Ep=350, Eg1=-18, Eg2=300, mu=171.8, rp=28.5k, Rs=47k, Rf=470k, RL=5k
From the equations:
Gain=6.89, Input impedance=67.2k, Output impedance=1.5k
From the SPICE simulation:
Gain=5.2, Input impedance=84.1k, Output impedance=1.25k
TIA,
Jaz
I'm still trying to get my head around the plate-to-grid feedback, and I went over Aiken's derivation using the 12AX7 as an example, the results matched pretty up well with the SPICE simulations. However, when using output pentode plate-to-grid feedback, the results no longer match, perhaps I am misapplying the equations or doing something stupid... Could someone please take a look and tell me what I did wrong... All I did was to look up the mu and rp for the pentode (6L6) and plug those figures into Aiken's equations, then compare them to the results from the SPICE simulation as shown below:
The operating parameters are:
Ep=350, Eg1=-18, Eg2=300, mu=171.8, rp=28.5k, Rs=47k, Rf=470k, RL=5k
From the equations:
Gain=6.89, Input impedance=67.2k, Output impedance=1.5k
From the SPICE simulation:
Gain=5.2, Input impedance=84.1k, Output impedance=1.25k
TIA,
Jaz
mu and ra figures for pentodes are notoriously unreliable. For small signal analysis it is more usual to consider it as a current source. The open loop gain becomes gm*Rl where Rl is the plate load resistor. If you substitute that for A in the equations and you should get much closer answers. Don't forget that gm needs to be the value at the operating point.
Cheers
Ian
Cheers
Ian
That open loop gain is small for a pentode. It is mainly due to the very small plate load of 5K. Are you sure about the 6mA/V for gm. Seems rather high to me. If the open loop gain is 30 then the 470K looks like 470K/30 = 15K6 at the input so the input impedance is 47K + 15K6 = 62K6
If spice thinks the input Z is about 84K that means the 470K must look like 17K at the input which implies the open loop is around 13 which implies spice thinks that gm at the operating point is lower than you have assumed.
Cheers
Ian
The thing to take away from this discussion is that both spice and the equations are only approximations. Add in the fact that real tube parameters vary by around 10% and you can see that both spice and the equations will really only get in the right basic region. Don't expect great accuracy from them. Once you are in roughly the area you want to be the next thing to do is build the circuit and find out what it really does.
Cheers
Ian
Cheers
Ian
Your points are well taken, nothing beats actually breadboarding and measuring the designs. As I am moving at the moment, all my tools/parts are packed away, so these days I am only able to read and sim ;-) But getting the theoretical stuff down is very important, that's why I appreciate your writeup on NFB - without knowing how things work and just running sims all day is not a good use of one's time for sure. OTOH, improving the SPICE models is a worthwhile activity, again, it is about knowing how things really work.
When I started working with 6P41C tubes I ran some sims that gave me what seemed to be strange results (I didn't read the datasheet closely enough) . The equations were generated from tube plate curves and the program paint_kip.jar (Dimitre ??sp). I figured I botched someting because of the sim results.
When I built the amp it was so close compared to the sim reasults that I was truly surprised.
Sim's allow us to do things with the tubes that would possibly damage them if we built the circuit, and when properly used can give very accurate results.
Then again, there is nothing like building an amp and listening to it.
When I built the amp it was so close compared to the sim reasults that I was truly surprised.
Sim's allow us to do things with the tubes that would possibly damage them if we built the circuit, and when properly used can give very accurate results.
Then again, there is nothing like building an amp and listening to it.
I agree, other than the parasitic elements from the layout and/or wiring which cannot be factored in the SPICE models, in general, the simulations can give fair good results provided that the models themselves are up to snuff (which unfortunately often are not the case). I think it is a good tool if used properly which I forget to do sometimes ;-)
BTW, has anyone reading this thread seen actual test results for the RH amplifiers? I know Mr. Kitic refuses to provide such information and prefers instead to rely on his published simulation results (which looks great), but surely with so many DIY builds out there, someone have taken some measurements?
Jaz
BTW, has anyone reading this thread seen actual test results for the RH amplifiers? I know Mr. Kitic refuses to provide such information and prefers instead to rely on his published simulation results (which looks great), but surely with so many DIY builds out there, someone have taken some measurements?
Jaz
Sorry if all this has been covered before, anyway, I have only recently started to read about Schade feedback and RH amps, and I tried to sift through some of the threads here and there, but I have not yet come across actual test results, such as FFT or THD plots, you know the standard stuff... Some links would be much appreciated.
Try looking at th "Why Pentodes " thread for a little detail. Most of this stuff was discussed back when the RH designs came out over five years ago. The threads should still be accessable.
The general conclusion was that Schade works best if the driver is of a fixed impedance - generally high - which suggests that Pentodes are the most appropriate driver for a Schade feedback amplifier. A FET buffer can also produce the same results, and a cascode triode driver approximates to a fixed impedance source.
Of all those options a Pentode driver is the most straightforward approach.
The Tubecad articles on Partial Feedback is probably one of the best descriptions of the principles and goes into why a Voltage to Current converter makes the best driver. Its a tricky read, but worth it.
Shoog
The general conclusion was that Schade works best if the driver is of a fixed impedance - generally high - which suggests that Pentodes are the most appropriate driver for a Schade feedback amplifier. A FET buffer can also produce the same results, and a cascode triode driver approximates to a fixed impedance source.
Of all those options a Pentode driver is the most straightforward approach.
The Tubecad articles on Partial Feedback is probably one of the best descriptions of the principles and goes into why a Voltage to Current converter makes the best driver. Its a tricky read, but worth it.
Shoog
Last edited:
I think the reason why a pentode makes a good driver for so-called 'Schade feedback' is that the driver needs to be good at driving a low impedance with low distortion. Most triodes rely on anode feedback for good linearity but this circuit deprives them of that. Pentodes have to be linear anyway as they cannot have that advantage, therefore they are better at driving into a low impedance.
Triodes are designed for linear mu, with varying degrees of success. Pentodes are designed for linear (maybe plus some square-law) transconductance. The latter is what is needed here.
The fact that pentodes have a high output impedance is irrelevant, so there is no reason to try other high impedance drivers. Linearity is the issue, not output impedance. A cascode will be even worse than a plain triode. To find a solution, first you have to understand the problem!
Triodes are designed for linear mu, with varying degrees of success. Pentodes are designed for linear (maybe plus some square-law) transconductance. The latter is what is needed here.
The fact that pentodes have a high output impedance is irrelevant, so there is no reason to try other high impedance drivers. Linearity is the issue, not output impedance. A cascode will be even worse than a plain triode. To find a solution, first you have to understand the problem!
One aspect of a triode driving a low Z Schade or shunt feedback network that is a bit confusing for analysis, is the triode's variable output impedance versus plate current.
A quick hand waving argument makes this look quite problematic, since the triode is going low Zp when the output tube is shutting off and going high Zout. (for linearity one would want these two to track rather than see-saw) Perhaps some light could be shed on whether this is an additional linearity problem, besides the restricted internal triode plate feedback (low Z loading), or is just another way of looking at the same issue.
For analysis, lets assume the triode driver is putting out X amount of voltage swing into ZLoad, and then look back at the linearity of its grid swing versus a pentode driver with similar gm and Zload and X voltage swing. The pentode looks like a triode with a frozen plate voltage (ie, screen V) versus the limited plate swing the triode is performing. One would expect the triode to be linearizing just a little better with that internal feedback, but maybe not, due to it's varying plate Z. Seems that one only needs to consider the current generated at the plate to predict the output V into the load Z, and the triode plate Z does not enter the calculation. ? (Let's assume both triode and pentode have the same un-bypassed cathode resistor as well.)
Any thoughts on this?
A quick hand waving argument makes this look quite problematic, since the triode is going low Zp when the output tube is shutting off and going high Zout. (for linearity one would want these two to track rather than see-saw) Perhaps some light could be shed on whether this is an additional linearity problem, besides the restricted internal triode plate feedback (low Z loading), or is just another way of looking at the same issue.
For analysis, lets assume the triode driver is putting out X amount of voltage swing into ZLoad, and then look back at the linearity of its grid swing versus a pentode driver with similar gm and Zload and X voltage swing. The pentode looks like a triode with a frozen plate voltage (ie, screen V) versus the limited plate swing the triode is performing. One would expect the triode to be linearizing just a little better with that internal feedback, but maybe not, due to it's varying plate Z. Seems that one only needs to consider the current generated at the plate to predict the output V into the load Z, and the triode plate Z does not enter the calculation. ? (Let's assume both triode and pentode have the same un-bypassed cathode resistor as well.)
Any thoughts on this?
Last edited:
Yes, a pentode can be thought of as a triode with constant anode voltage but it may not be quite as simple as that. As I said, an triode is optimised for constant mu while a pentode is optimised for constant gm - assuming, that is, that either have actually been optimised. The latter is what is needed for a 'Schade' driver.
Thanks for the tip, I think this is the one you were referring to, got some good design examples and even some listening impressions...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.