no matter how i moved the working point,still high distortion
833 was a big surprise-positive; but 100w heaters kinda sucks
833 was a big surprise-positive; but 100w heaters kinda sucks
OK
Please take into account that the simulator is based on manufacturers' anode characteristics of tubes (ideal or averaged characteristics), not on real tubes measurements.
There remains the question of how exactly they are mapped. You can check this by comparing selected points given by the simulator to the set of points of the data read from the catalog. If there are any significant spreads, I'll correct them.
Available manufacturer's data for large tubes, for example (GM100, GM70 etc) are very poor quality (I mean that anode characteristics are very small and it is very difficult to read anode current/anode voltage values). This may be the reason of the problem of accuracy. If someone has good looking characteristics of these tubes I'll be happy to use them ;-).
Please also note that calculations are performed using values calculated on the basis of the mathematical model, which is also somehow approximate.
And one more thing, in reality the THD measurements are in fact THD + N values and and simulator calculates only THD and only for 5 first frequency components. So if your real amplifier generates noise and more harmonics, measured values will be greater than calculated.
12AX7 = ECC83, which is in there.
5965 isn't designed for audio use and can be very microphonic. They're actually intended for use in tube-based computers.
Just took a quick look in my tubebox and misstook the 5965 for the 5751,
no idea why and from where I got those tubes, sorry.
Not shure how much the 5751 differs from the 12AX7/ECC83 but the
comparison is interessting.
So never mind about the 12ax7 😱
Still, the 6cg7 and the 5751 🙂
Maybe somewhere in between the 6sn7 and 6k5gt.
I get compliments nearly daily on a 2 stage amp I recently made for someone. He likes the 6K5GT as the preamp driving a 6L6GC.
I can confirm that the gm100 is off by a considerable margin. With 1650v on plate at 0v bias and 17c on filaments i get 165mA on average depending on tube
Inviato dal mio GT-I9301I utilizzando Tapatalk
Inviato dal mio GT-I9301I utilizzando Tapatalk
But this is not the OP fault's (who I thank greatly). The curves are crap.
Inviato dal mio GT-I9301I utilizzando Tapatalk
Inviato dal mio GT-I9301I utilizzando Tapatalk
Maybe, I'm not really a "tubeguy", so I'm not that familiar with what's close
enough to some other tube. I do make some nice speakers and I like the sound
of tubes 🙂
Trying to learn as I go along though 😱
Probably someone already said this but I think that people are getting overly stressed on the accuracy of and choice of tubes in the simulator.
It is an excellent instrument but I believe it should serve as a guideline. Nothing can replace actually in-circuit testing. Yes, it is a nuisance and yes, it will take up 80% of the build time but it is a necessary part of the game. The curves on the gm100 datasheet are terrible yet the tube performs well for audio. No one can predict that with a simulator...
Inviato dal mio GT-I9301I utilizzando Tapatalk
It is an excellent instrument but I believe it should serve as a guideline. Nothing can replace actually in-circuit testing. Yes, it is a nuisance and yes, it will take up 80% of the build time but it is a necessary part of the game. The curves on the gm100 datasheet are terrible yet the tube performs well for audio. No one can predict that with a simulator...
Inviato dal mio GT-I9301I utilizzando Tapatalk
The million dollar question is what is model based on... which is "suppose" to be the data sheet curves...
Even in the ideal case where you have perfectly modeled the "Bogie" valve in simulation, you mileage will vary significantly because the actual real valves you are using rarely represent the data sheet curves... just normal variance...
All the Major tube manufacturers back in the day "cheated" a bit ...they cherry picked a perfect tube from production, then when they transferred the Negative transparencies of the curves they touched them up with french curves to make them look a bit nicer...before making the plates for printing the data sheets...
There are methods to create very accurate models, but so far I have not seen them used in the tube models..
Even in the ideal case where you have perfectly modeled the "Bogie" valve in simulation, you mileage will vary significantly because the actual real valves you are using rarely represent the data sheet curves... just normal variance...
All the Major tube manufacturers back in the day "cheated" a bit ...they cherry picked a perfect tube from production, then when they transferred the Negative transparencies of the curves they touched them up with french curves to make them look a bit nicer...before making the plates for printing the data sheets...
There are methods to create very accurate models, but so far I have not seen them used in the tube models..
So trying to learn here 🙂
Simulated the 6n6p.
Ra = Ri x 3
Voltage drop 1/2 to 2/3 of B+ (300V)
Distortion at a decreasing slope. It's possible to lower the total amount but then H3 goes up.
Comments ?
Simulated the 6n6p.
Ra = Ri x 3
Voltage drop 1/2 to 2/3 of B+ (300V)
Distortion at a decreasing slope. It's possible to lower the total amount but then H3 goes up.
Comments ?
It is precisely the reason I gave (some time ago) a link to the extended version of the simulator, which I called tube optimizer.
Here is the direct link:
http://www.trioda.com/tools/tube.html
The program in the lower part of the page allows you to enter the measurement data for a specific REAL tube (exemplary data is entered into a text field - you can decode the data format from it). By using an evolutionary algorithm (simulated annealing algorithm) by successive pressing [Optmize] button and observing the curves aproximating your measerment points you can find the model parameters of your "personal" tube and import them to the Simulator (upper part of the page). And then simulate at will 🙂
Take into account that annealing algorithm, like other evolutionary algorithms not always finds the best data aproximation. If it can't find the model you have to repeat the simulation from the beginning etc. I am working now to improve the optimization method.
Last edited: