The economics and component choice changes with DIY, but basic physics of what reduces distortion applies universally.
Re your 2nd point, fair enough, but if you prefer the sound of triodes, and I agree that many do, then the supposed benefit of UL on SE re distortion has no relavence.
I certainly aggree that using the secondary for feedback to the cathode works well - in reducing distortion of course, taking you away from tube sound.
Your preconception seems to be that your mathematics is the only thruth and all that deviates from that is false or tests made wrong ??
When I told that I had measured the characteristics of the SE output stages, I meant that. There were only the tube, OPT, bias circuit, power supplies, resistive load and the signal generator (able to supply > 100 Vpp with 0,03 % THD). No preceeding stages. THD were analyzed with HP8903B etc. etc.
What else do you want to know about my test method ?
Can we stick to what the OP is asking? He clearly doesn't want to engineer an amp, and he's asking a simple question, will replacing the 12AX7 with a 12AT7 work?
When test results differ from well established and proven theory, it is reasonable and sensible to query the test method.
However, since your SPICE data is down to the equivalent of 0.2% distortion, and your oscillator is good to 0.03% there should be no problem there. You didn't say what it was though, so if there are any quirks we can't know. The HP8903B is most certainly up to the task. Perhaps there was another problem, although with only the output stage I must admit there's not much to trick you. Output transformer?
I note that you gave SPICE results, but your wording does not actually state numerical test data, only that test results were similar. How similar is your "similar"?
Last edited:
Since I have done my tests with different setups, then I must have a bunch of "faulty" output transformers🙂.
I live part of the year in Spain and now my documents are at home in Finland, but in the cases I remember when I have tested SE output stages,
the general behaviour is that UL gives lower THD (and higher output power).
This is exactly identical as LTspice shows the SE circuit to behave.
I can show my LTspice simulations if that is what you want to see.
Maybe it would now be time to you to take the soldering iron on hand and begin to study how the things are ?
Last edited:
I made some more simulations of SE output stage with adjustable amount of UL.
I begun with triode connected stage as a reference and continued with 10 % steps, i.e. UL 90%, UL80%... etc.
The last test was done with pentode circuit (0 % UL).
The used power level was 2 W.
In all cases I recorded THD, 2nd and 3rd harmonic levels.
The used circuit was as attached.
The THD decreased quite straightforward from 1,06 % (triode) to 0,56 % (UL 10 %) and then increased again to 0,90 % with pentode.
2nd harmonic followed the same route, it decreased from -40,0 dB to -49,3 dB until UL10%. Then with pentode (UL 0%) it increased to -45,5 dB
3rd harmonic increased quite linearly from -78 dB (triode) to -42,9 dB (pentode).
At the lowest point of THD 2nd and 3rd harmonics had quite indentical level, at UL 10%.
The optimum UL-percentage was surprisingly low, but this maybe due to used 6P36S (EL504) sweep tube that has very effective screen grid.
I begun with triode connected stage as a reference and continued with 10 % steps, i.e. UL 90%, UL80%... etc.
The last test was done with pentode circuit (0 % UL).
The used power level was 2 W.
In all cases I recorded THD, 2nd and 3rd harmonic levels.
The used circuit was as attached.
The THD decreased quite straightforward from 1,06 % (triode) to 0,56 % (UL 10 %) and then increased again to 0,90 % with pentode.
2nd harmonic followed the same route, it decreased from -40,0 dB to -49,3 dB until UL10%. Then with pentode (UL 0%) it increased to -45,5 dB
3rd harmonic increased quite linearly from -78 dB (triode) to -42,9 dB (pentode).
At the lowest point of THD 2nd and 3rd harmonics had quite indentical level, at UL 10%.
The optimum UL-percentage was surprisingly low, but this maybe due to used 6P36S (EL504) sweep tube that has very effective screen grid.
Attachments
You can use this schem with 2 parallel KT88 for more power with Yours OTs.
Schem is very easy to make it....
So, your actual test results are not available. How very convenient.
As I said before, it is unwise to rely on SPICE (whether LTSpice or another incantation) to predict distortion levels. The problem isn't with SPICE it is that the models don't accurately model the curvature of the grid characteristics, especially of the screen grid characteristic, which for most tubes is more curved than it is for the control grid. There's a certain amount of luck here unless you are intimately familiar with the actual model you are using for the particular tube type, as there are various models around.
Having said that, there seems to be an error in your simulation circuit. I've not used LTSpice, as I have access to a commercial version. I assume the device shown as "E1" is a VCVS with a gain of 0.5 which you have assumed simulates a 50% of turns transformer tapping. This is incorrect. The screen in UL doesn't just act as an input for feedback purposes - it also delivers a small amount of power to the transformer. Not much, typically around 5% of the total, but it contains plenty of even order harmonic. To more accurately simulate UL, you should model it with a three-winding transformer. And the tube model you are using must corectly simulate the control grid/screen grid gm as well as the screen/anode gm. Not just as a given value but with an apparopriate power series. Does the model you are using do this?
Also note a convention that dates back to when UL was invented. When authors talk about the transformer tapping percentage for the screen connections, the percentage quoted relates to the impedance. So what is called a "50%" tap would actually a 75% turns or voltage tap (which is too high of course). It does vary with different authors - some do quote the turns percentage.
Haffler and Keroes, and the applications lab reports I have (all tested with a multi-tapped transformer and push pull [naturally]) all show 3rd harmonic decreasing from the pure pentode level quite rapidly down to a minimum, not much different to the triode level, at around the 30 to 50% tapping point depending on the tube used and other factors, the following a more or less straight line as the tappings are moved towards the anode connections
Last edited:
My actual tests with several different pentodes/beam tetrodes show that SE UL will give lower THD than SE triode connection.
Concerning the LTspice simulation results I think it is not very likely that all/most pentode models available would be incorrect in such a way that the simulation results give completely opposite information than your original statement:
If the even order distortion, which is already dominating component in triode SE, were increased when UL SE is used, then the THD would be bigger in UL, but it is not. It is lower.
The simulation circuit I showed above was done just because the adjustment of UL was easy, but it gives similar 2nd order harmonic behaviour than the "real" that I have been using with normal simulations. (below).
I think that this discussion does not proceed before you build a test circuit and get the results with it.
Concerning the LTspice simulation results I think it is not very likely that all/most pentode models available would be incorrect in such a way that the simulation results give completely opposite information than your original statement:
If the even order distortion, which is already dominating component in triode SE, were increased when UL SE is used, then the THD would be bigger in UL, but it is not. It is lower.
The simulation circuit I showed above was done just because the adjustment of UL was easy, but it gives similar 2nd order harmonic behaviour than the "real" that I have been using with normal simulations. (below).
I think that this discussion does not proceed before you build a test circuit and get the results with it.
Attachments
Last edited:
Let me sum this up:-
1) You claimed at the start that practical tests backed up the conclusions you drew from SPICE testing but..
2) You only gave SPICE result data, and no test data.
3) You did not explain how you tested until pressed more than once.
4) It turned out the test results are not available.
5) You used an incorrect SPICE simulation, and AFTER this was pointed out, suddenly anounced a correct version without any result data.
Why was the correct sim not posted first?
While your logic in your (b) above is sound, maybe you should stop making claims until you are ready to present the complete picture, and you have authentic data upon which to exercise your thinking.
Don't stick to non-essential.
The first sim schematic I presented shows why I showd it first. With it was easy to have different UL-percentages, much easier and faster than the "correct" one (that I have used for years...).
Let me sum this up:
- what concrete do you have to support your claim ? (except this: sin(2x) = 1 - 2sin(x)^2)
As I said earlier, this discussion does not proceed before you build a test circuit and get the results with it.
If you have a soldering or iron and you can even use it, just go ahead.
- Status
- Not open for further replies.