Hi all. I've been working on a design for an OTL tube headphone amp. As my starting point, I used the design created by @indaco and @artosalo that can be found here: https://www.diyaudio.com/community/threads/the-6n13s-artosalo-otl-headphone-amplifier.321470/
I have some nice 12au7 and KT66 tubes on hand and I like the sound of them, so I thought I'd try adjusting the design to use those tubes and modeling it in LTSpice (also as an exercise in LTSpice.
I made the following changes to the design:
I set the sine wave input an amplitude of 2.82 to represent the common 2Vrms output from a DAC and put a potentiometer there with the wiper at 0.75.
The result is here:
The result is significantly more power at a lower distortion than my model of the Bottlehead Crack. Turned all the way up, this does 360mW into 300 ohms.
I've heard some people say the 6SN7 is better than the 12au7, so I made a model with those as well and readjusted the resistor values. This version has more gain with concurrent slightly higher distortion than the 12au7 version I came up with:
I do have nice 12au7s on hand, but if I'm going to go to the trouble of building an amp, maybe I want the better tube? This version will do 500mw into 300 ohms
Output impedance is about 85 ohms, which is fine for the high impedance dynamic headphones I intend to use this with, and again much better than the Crack. (Please no suggestions on how to lower the output impedance, this isn't my goal.)
I've included both simulations here for anyone who wants to play with them.
Questions:
Any other comments and suggestions appreciated!
There's a lot of other things I need to figure out before I could actually build this, in particular the power supply and heater wiring and voltages. The original Indaco/Artosalo design used an SMPS for the power heater; I'd like to avoid that, but I need to research transformers and such. I also need to figure out the general wiring and assembly--how I'd actually put this together assuming I'd use point-to-point wiring (which I've never done!). But right now I'm just focused on perfecting the gain stage. I love OTL headphone amps and this has the potential to be a dream piece of gear for me.
Edit: made a mistake on the potentiometer setting which I have corrected
I have some nice 12au7 and KT66 tubes on hand and I like the sound of them, so I thought I'd try adjusting the design to use those tubes and modeling it in LTSpice (also as an exercise in LTSpice.
I made the following changes to the design:
- Changed tubes to 12au7 and KT66
- Ran the control grid of the KT66 back to the B+/Anode line with a small resistance as is commonly done
- Raised the B+ voltage to 300V
- Adjusted the resistor values for the new tubes, with an aim towards minimizing distortion
- Removed the input capacitor—since both tubes are cathode biased, it should be unnecessary. Replaced with a grid-stopper resistor
- Added a potentiometer
I set the sine wave input an amplitude of 2.82 to represent the common 2Vrms output from a DAC and put a potentiometer there with the wiper at 0.75.
The result is here:
The result is significantly more power at a lower distortion than my model of the Bottlehead Crack. Turned all the way up, this does 360mW into 300 ohms.
I've heard some people say the 6SN7 is better than the 12au7, so I made a model with those as well and readjusted the resistor values. This version has more gain with concurrent slightly higher distortion than the 12au7 version I came up with:
I do have nice 12au7s on hand, but if I'm going to go to the trouble of building an amp, maybe I want the better tube? This version will do 500mw into 300 ohms
Output impedance is about 85 ohms, which is fine for the high impedance dynamic headphones I intend to use this with, and again much better than the Crack. (Please no suggestions on how to lower the output impedance, this isn't my goal.)
I've included both simulations here for anyone who wants to play with them.
Questions:
- 12au7 vs 6SN7 in this amp?
- Any suggestions for modifications? Tweaks? I'm particularly interested in lowering that 3rd order harmonic distortion a little (though I'm not sure how audible it actually is? Maybe not at all?) (without removing the KT66 from the design--yes, I know beam tetrodes are more likely to produce 3rd order harmonic distortion, the idea here is to use the KT66)
- I think about how the Feliks amplifiers use two dual triodes in the input stage. I'm not sure how they have them configured--there's a million ways to do it--but I'm wondering if adding a second input triode somewhere would improve the design
Any other comments and suggestions appreciated!
There's a lot of other things I need to figure out before I could actually build this, in particular the power supply and heater wiring and voltages. The original Indaco/Artosalo design used an SMPS for the power heater; I'd like to avoid that, but I need to research transformers and such. I also need to figure out the general wiring and assembly--how I'd actually put this together assuming I'd use point-to-point wiring (which I've never done!). But right now I'm just focused on perfecting the gain stage. I love OTL headphone amps and this has the potential to be a dream piece of gear for me.
Edit: made a mistake on the potentiometer setting which I have corrected
Last edited:
Looks promising, I think I can think of a few tweaks.
Could you attach your include .inc files so we can follow along?
Could you attach your include .inc files so we can follow along?
Even the low gain 12AU7 is a little high, I noticed you are throwing away some gain with R3. Try removing C2.
How much gain are you looking for? and are you apposed to negative feedback in your circuit?
How much gain are you looking for? and are you apposed to negative feedback in your circuit?
Last edited:
@mr_zener So removing C2 seems to pretty significantly increase distortion:
The Bottlehead Crackhas 15db of gain, so I consider that kind of a baseline, but I'm open suggestion as long as I can run my Sennheisers and ZMFs to a reasonable volume. R3, assuming I'm understanding correctly, is what does the cathode biasing for the KT66. Removing that sends distortion into the stratosphere:
I don't mind playing with negative feedback, especially local negative feedback. I think ultimately I want the design not to have global negative feedback, but as I said I'm willing to play with it and see what it does to the simulation.
The Bottlehead Crackhas 15db of gain, so I consider that kind of a baseline, but I'm open suggestion as long as I can run my Sennheisers and ZMFs to a reasonable volume. R3, assuming I'm understanding correctly, is what does the cathode biasing for the KT66. Removing that sends distortion into the stratosphere:
I don't mind playing with negative feedback, especially local negative feedback. I think ultimately I want the design not to have global negative feedback, but as I said I'm willing to play with it and see what it does to the simulation.
LOL that was with the potentiometer turned up to 75%. Here's the apples-to-apples 25% version:
Also the ASC file
Also the ASC file
(Sorry to keep spamming this thread)
So while I do have to turn the volume knob up to 75% to get the same output as I do at 25% in the first design, look at the distortion profile when I do:
Like that's ridiculous.
These resistor values btw are based on the recommendations in the original tube manuals rather than math, which I used for the original.
So while I do have to turn the volume knob up to 75% to get the same output as I do at 25% in the first design, look at the distortion profile when I do:
Like that's ridiculous.
These resistor values btw are based on the recommendations in the original tube manuals rather than math, which I used for the original.
On the right track but R4 is far to low and will far exceed KT66 max current. Have a look at your current draw.
Here is a link for the KT66 datasheet: http://www.r-type.org/exhib/aaj0183.htm
I also seem to be getting far different results than yours, could you also change your potentiometer to a resistor divider for the time being.
Here is a link for the KT66 datasheet: http://www.r-type.org/exhib/aaj0183.htm
I also seem to be getting far different results than yours, could you also change your potentiometer to a resistor divider for the time being.
So I got that value from the KT66 Data Sheet (attached):
That's probably also for Grounded Cathode configuration rather than Cathode Follower like in my design. That said in the link you provided it states the maximum current as 85mA, however in the document I've attached it says the maximum anode current is 200 (this design currently does about 168mA in the simulation). From the document attached to the webpage you linked to, it's not clear to me that the 85mA refers to maximum current as opposed to typical current:
So I'm not sure this is correct. I may well be wrong, though, and please let me know if I am.
That's probably also for Grounded Cathode configuration rather than Cathode Follower like in my design. That said in the link you provided it states the maximum current as 85mA, however in the document I've attached it says the maximum anode current is 200 (this design currently does about 168mA in the simulation). From the document attached to the webpage you linked to, it's not clear to me that the 85mA refers to maximum current as opposed to typical current:
So I'm not sure this is correct. I may well be wrong, though, and please let me know if I am.
What bias voltage are you getting? K-G1? The voltages aren't noted on the schematic. You don't have a cap coupling the grid.
@20to20 oh, right. It's about 51V between R9 and the grid of the KT66. (I'm trying to figure out how to automatically display the voltages on the nodes, but it doesn't seem to be done as easily on the Mac as on Windows.) You're right that I don't have a coupling capacitor there, or even a grid-stopper resistor. (I just tried a grid-stopper there, and it made the distortion very slightly worse.)
However, neither does the original design--I'd say the cathode biasing on R3 makes it unnecessary, but the Bottlehead Crack design doesn't even have that:
¯\(ツ)/¯
However, neither does the original design--I'd say the cathode biasing on R3 makes it unnecessary, but the Bottlehead Crack design doesn't even have that:
¯\(ツ)/¯
To find the bias voltage you would go directly from grid to cathode with your meter probes or just do one at at a time and subtract one from the other for the difference. The grid voltage must be lower than the cathode.
Oh, I see. Thanks for the explanation. Cathode is about 54 while the grid is about 51:
So the bias would be -3V, yes?
So the bias would be -3V, yes?
That would be it. So then you can find the plate current on the triode chart. ....seems to be off the chart.
... you're right! It's somewhere up here!
It fits fine on the Tetrode chart, but I am using the tube in a triode configuration. Huh. Not sure what that means--it seems like the tube can handle the current because it's shown on the Tetrode chart and beneath the stated maximum, but that it's not standard (optimal?) triode behavior?
Like I'm wondering if it's one of those things where it looks great in the simulator but would blow up on the bench.
It fits fine on the Tetrode chart, but I am using the tube in a triode configuration. Huh. Not sure what that means--it seems like the tube can handle the current because it's shown on the Tetrode chart and beneath the stated maximum, but that it's not standard (optimal?) triode behavior?
Like I'm wondering if it's one of those things where it looks great in the simulator but would blow up on the bench.
Last edited:
In my findings the KT66 is not the best output tube for the job, due to output impedance being relatively high, that about half's the incoming signal (so about 300Ohm). Check by increasing the load.
Maybe consider another output tube, like 6AS7 / 6080 / 6BL7?. they are also duel triodes so you could use them in parallel, 1 tube per channel (would be excellent if not a little over-kill they are power hungry filaments for the 6AS7 and 6080 at 6.3V @ 2.5A).
I have a simulation that would probably work OK, but without feedback and a more complex design, I'm not sure you can do much better.
It all depends what your after and if you want to switch output tubes.
Maybe consider another output tube, like 6AS7 / 6080 / 6BL7?. they are also duel triodes so you could use them in parallel, 1 tube per channel (would be excellent if not a little over-kill they are power hungry filaments for the 6AS7 and 6080 at 6.3V @ 2.5A).
I have a simulation that would probably work OK, but without feedback and a more complex design, I'm not sure you can do much better.
It all depends what your after and if you want to switch output tubes.
Oh, wait. I see that the screen resistor is 3K, that should put it on the pentode chart. You had me fooled looking at the triode data. Must take another look but -3v is the bias.