If you continue to have problems getting that 10M45S model to work, maybe you can try this DN2540 model instead.
Close LTspice, then.,..
Copy and paste that into your standard.mos file in \LTspiceXVII\lib\cmp\ or wherever you store your standard.mos file.
Then when you reopen LTspice use the device picker to insert an nmos symbol into your schematic, then right-click on the nmos symbol and choose DN2540.
Code:
.model DN2540 VDMOS (Rg=1 Vto={-1.52-4.5m*(temp-25)} Lambda=6m Rs={0.35*(1+3m*(temp-25))} Kp={800m/(1+6m*(temp-25))} Ksubthres=0.11 Mtriode=0.35 Rd={6*(1+3m*(Temp-25))} Cgdmax=10p Cgdmin=1p a=0.25 Cgs=100p Cjo=200p m=0.7 VJ=2.5 IS=4.0E-8 N=2.4 Rb=10 )Close LTspice, then.,..
Copy and paste that into your standard.mos file in \LTspiceXVII\lib\cmp\ or wherever you store your standard.mos file.
Then when you reopen LTspice use the device picker to insert an nmos symbol into your schematic, then right-click on the nmos symbol and choose DN2540.
Thanks guys!
So set up this way:
I get basically identical results in terms of anode voltage/current, gain, and distortion. Would this make an improvement in sound quality on the real build? No idea. Though it'd be an easy thing to experiment with once I had it on the bench.
So set up this way:
I get basically identical results in terms of anode voltage/current, gain, and distortion. Would this make an improvement in sound quality on the real build? No idea. Though it'd be an easy thing to experiment with once I had it on the bench.
Attachments
Might need to add a diode. Honestly, still feeling my way this stuff and throwing things at the wall... which is a way of saying I barely know what I'm doing and appreciate everyone's patience.
This project falls into the category of an idea that could eventaully be made to ''work'', but is a bad idea. The patience that most folks have on the board comes from lack of experience and having the curiosity to watch someone work out a plan to see for themselves what can fly or fail. Using a KT66 for the output tube of a headphone amp seems like a really cool tube to look at while listening to music and having some kind of conversation piece to show off but it's not the tube to use in this type of device for the purpose. So, you'll find the patience here could let you continue to walk farther down a path that just leads to dissapointment and wasted time. What do you suppose the distortion figure might come in at when you finalize this circuit? What you are building is a HIFI amp with no FB and taking 10% of it's power. Unfortunately, there is no shortcut, including sims, to learning this stuff. Find a good already proven circuit and study that and 10 more. Ok, good luck, 20, oao.
I have to agree with a lot of the last post from 20to20.
There are plenty of proven designs for OTL tube headphone amps. Most use a 6AS7G as the output tube, configured as a cathode follower. There are a few commercial products like this as well.
Examples of commercial 6AS7G amps:
Bottlehead Crack
XDUOO TX-66
Very nice looking DIY headphone amp using 6AS7G
John Broskie idea for an 'Aikido' 6AS7G based headphone amplifier for driving 50 ohm impedance cans
Fully documented 6AS7G based headphone amp from AudioXpress
A big list of DIY vacuum tube headphone amp ideas
6AS7G is a big twin triode with very low mu (gain) and lots of power capability (plate dissipation of 13W per triode section).
6080 is the more modern version with the same specs.
There are some Russian equivalents too.
The reason there are so many tube headphone amps made with this basic topology is that it makes good sense within the hard limitations of vacuum tubes. You need a low output impedance to drive headphones (at least by tube standards), and if you don't use an output transformer then you'll need to use a big, honkin' cathode follower to get that low Zout.
Just like with speaker amplifiers using output tubes, an output transformer takes the relatively high output impedance of vacuum tubes and steps it down to provide a very low output impedance to drive 4 or 8 ohm speakers. This same idea holds true for 32 ohm headphones. For instance, the output impedance of a triode-wired EL84 is about 2k ohms, which needs to be stepped down a lot to drive 32 ohm headphones. You want the Zout of your headphone amp to be lower than half the headphone's impedance -- and preferably much lower than that (like at least 5X lower than the headphone's impedance). You also want to achieve that with decent power output, wide enough bandwidth, flat frequency response, good transient response (no oscillations, resonances), and low distortion. That's the challenge.
There are plenty of proven designs for OTL tube headphone amps. Most use a 6AS7G as the output tube, configured as a cathode follower. There are a few commercial products like this as well.
Examples of commercial 6AS7G amps:
Bottlehead Crack
XDUOO TX-66
Very nice looking DIY headphone amp using 6AS7G
John Broskie idea for an 'Aikido' 6AS7G based headphone amplifier for driving 50 ohm impedance cans
Fully documented 6AS7G based headphone amp from AudioXpress
A big list of DIY vacuum tube headphone amp ideas
6AS7G is a big twin triode with very low mu (gain) and lots of power capability (plate dissipation of 13W per triode section).
6080 is the more modern version with the same specs.
There are some Russian equivalents too.
The reason there are so many tube headphone amps made with this basic topology is that it makes good sense within the hard limitations of vacuum tubes. You need a low output impedance to drive headphones (at least by tube standards), and if you don't use an output transformer then you'll need to use a big, honkin' cathode follower to get that low Zout.
Just like with speaker amplifiers using output tubes, an output transformer takes the relatively high output impedance of vacuum tubes and steps it down to provide a very low output impedance to drive 4 or 8 ohm speakers. This same idea holds true for 32 ohm headphones. For instance, the output impedance of a triode-wired EL84 is about 2k ohms, which needs to be stepped down a lot to drive 32 ohm headphones. You want the Zout of your headphone amp to be lower than half the headphone's impedance -- and preferably much lower than that (like at least 5X lower than the headphone's impedance). You also want to achieve that with decent power output, wide enough bandwidth, flat frequency response, good transient response (no oscillations, resonances), and low distortion. That's the challenge.
Last edited:
To @20to20 so this project was initially inspired by people using the KT66 in the OTL Feliks Euforia. Donald North Audio also uses a KT66 in a headphone amp, though that one is transformer-coupled. I appreciate the admonition to study more circuits but I don’t really understand the objection to this one. Why would the distortion profile be meaningfully different than the simulation?
To @rongon as I said in my first post this is explicitly for high impedance dynamic headphones, I have no interest in 32 ohm headphones for this project. In the simulation I get 95 ohms output impedance which is perfect for my headphones (and less than the Crack)
The thing that’s interesting about this project is to do KT66 tubes. If I wanted 6AS7, I’d just build another Crack.
To @rongon as I said in my first post this is explicitly for high impedance dynamic headphones, I have no interest in 32 ohm headphones for this project. In the simulation I get 95 ohms output impedance which is perfect for my headphones (and less than the Crack)
The thing that’s interesting about this project is to do KT66 tubes. If I wanted 6AS7, I’d just build another Crack.
Thanks for your reply w/ explanations.
I hope you don't mind answering a few more questions about the proposed circuit:
1. I've never seen a DN2540 current sink wired up the way you have it. What is the purpose of R13 (380 ohms)?
2. Why is the gate of the DN2540 connected to the cathode of the KT66? Why not connect the depletion mode MOSFET's gate to signal ground, as is normally done?
3. What is the purpose of R1 (2.9k) in the plate of the KT66?
SPICE can return fictitious results if fed data that it isn't programmed to recognize. In other words, there are circuits that model OK in SPICE that will not work in real life. The designer needs to understand the real world limitations of real world devices. For example, you can set your B+ to 600V and your KT66's plate+screen grid current to 100mA and get some spectacular distortion numbers, but in real life the tube will melt under those conditions.
You might want to breadboard your design using used tubes (cheap) and take DC voltage measurements to make sure things work as expected.
Thanks.
I hope you don't mind answering a few more questions about the proposed circuit:
1. I've never seen a DN2540 current sink wired up the way you have it. What is the purpose of R13 (380 ohms)?
2. Why is the gate of the DN2540 connected to the cathode of the KT66? Why not connect the depletion mode MOSFET's gate to signal ground, as is normally done?
3. What is the purpose of R1 (2.9k) in the plate of the KT66?
SPICE can return fictitious results if fed data that it isn't programmed to recognize. In other words, there are circuits that model OK in SPICE that will not work in real life. The designer needs to understand the real world limitations of real world devices. For example, you can set your B+ to 600V and your KT66's plate+screen grid current to 100mA and get some spectacular distortion numbers, but in real life the tube will melt under those conditions.
You might want to breadboard your design using used tubes (cheap) and take DC voltage measurements to make sure things work as expected.
Thanks.
Last edited:
Hi @rongon ,
I wouldn't take the DN2540 version of the circuit too seriously; I was playing around with doing a constant current mod to imitate the Speedball mod on the Bottlehead Crack, and I tried wiring it up to the output in imitation of this design: https://www.diyaudio.com/community/...with-12au7-and-kt66-tubes.427541/post-8012530 and it measured better than wiring it to ground (in the simulation), but I'm not sure this is really the way to go, and as I say I'm not sure it would even have a discernible difference in the output. (If you scroll back you can see a different design with the transistor going to ground with a different transistor that blew up the tube in the sim.)
The R1 anode resistor is there to prevent the KT66 from melting down as I explained here: https://www.diyaudio.com/community/...au7-and-kt66-tubes.427541/page-3#post-8010300 -- without that resistor, the anode voltage and current surpasses the tube limits when you turn up the volume, but with that resistor you can turn the volume all the way up without that happening.
For reference this is my current "best" version of the design:
Which in the sim has 0.13% THD.
I wouldn't take the DN2540 version of the circuit too seriously; I was playing around with doing a constant current mod to imitate the Speedball mod on the Bottlehead Crack, and I tried wiring it up to the output in imitation of this design: https://www.diyaudio.com/community/...with-12au7-and-kt66-tubes.427541/post-8012530 and it measured better than wiring it to ground (in the simulation), but I'm not sure this is really the way to go, and as I say I'm not sure it would even have a discernible difference in the output. (If you scroll back you can see a different design with the transistor going to ground with a different transistor that blew up the tube in the sim.)
The R1 anode resistor is there to prevent the KT66 from melting down as I explained here: https://www.diyaudio.com/community/...au7-and-kt66-tubes.427541/page-3#post-8010300 -- without that resistor, the anode voltage and current surpasses the tube limits when you turn up the volume, but with that resistor you can turn the volume all the way up without that happening.
For reference this is my current "best" version of the design:
Which in the sim has 0.13% THD.
Some ideas for you to consider (some of these ideas may offer conflicting requirements):
- Try using a 6SN7 or 6FQ7 in the gain stage.
- Try running the output stage in triode mode.
- Try an output stage tube that is easier to drive, like a 7591, EL84, or 6V6, or even a low Rp preamp tube like a 6H30 or equivalent.
- Try for a solution that direct couples the gain stage to the output stage.
- Try for a solution that does not require the cathode of the gain stage to be bypassed.
OK, thanks again.
1. That R1 resistor is reducing the gm of your KT66, increasing Zout and worsening performance. It may be better to bias the KT66 cooler to prevent it from overheating at high output levels. It will still be in class A because the circuit is single-ended.
2. It looks like you wired the DN2540 incorrectly. A depletion mode MOSFET is a lot like a vacuum tube. In normal operation it needs to have its gate held negative to its source (like a tube needs its grid held negative to its cathode). The set resistor (Rset) is analagous to the cathode load resistor (Rk) for a self-biased vacuum tube.
If the CCS were to be used as a cathode load (current sink), the bottoms of R7 and R3 would go to signal ground. The drain of the MOSFET would connect to the cathode of the vacuum tube.
3. You wrote "Which in the sim has 0.13% THD." At what output level and into what load? Was that 1Vrms into 300 ohms? Or 1mW into 300 ohms?
I can't tell anything about those commercial headphone amp designs you mentioned because there are no schematics, etc. in the documentation.
1. That R1 resistor is reducing the gm of your KT66, increasing Zout and worsening performance. It may be better to bias the KT66 cooler to prevent it from overheating at high output levels. It will still be in class A because the circuit is single-ended.
2. It looks like you wired the DN2540 incorrectly. A depletion mode MOSFET is a lot like a vacuum tube. In normal operation it needs to have its gate held negative to its source (like a tube needs its grid held negative to its cathode). The set resistor (Rset) is analagous to the cathode load resistor (Rk) for a self-biased vacuum tube.
If the CCS were to be used as a cathode load (current sink), the bottoms of R7 and R3 would go to signal ground. The drain of the MOSFET would connect to the cathode of the vacuum tube.
3. You wrote "Which in the sim has 0.13% THD." At what output level and into what load? Was that 1Vrms into 300 ohms? Or 1mW into 300 ohms?
I can't tell anything about those commercial headphone amp designs you mentioned because there are no schematics, etc. in the documentation.
I think R6 is meant to represent the 300 ohm impedance headphones, so R5 is the standard pull-down resistor.
I think 510k is too high a value for R5. I'd choose something like 4.7k, since that won't add much to the 300 ohm load the amp is designed to drive.
I think 510k is too high a value for R5. I'd choose something like 4.7k, since that won't add much to the 300 ohm load the amp is designed to drive.
kward
Can you share your KT66 model?
I just whipped together a sim in LTspice using Adrian Immler's 6L6GC model, and got interesting results.
1. The 6L6GC plate-cathode voltage is only 170V.
2. The 6L6GC screen grid (g2) voltage is 270V.
3. The 6L6GC plate is only dissipating 7.5W. KT66 is rated at something around 20W plate dissipation (that's from memory, I'm probably off by a bit). You could use a 6V6GT there and get the same performance.
4. The THD I get at 1V rms output into 300 ohms is 0.38%, which is actually not too bad (which surprised me).
5. That 2.9k R12 resistor is burning hot, dissipating 4W all by itself. That seems to me to be a waste.
6. When you tested the circuit without R12 installed, how did you tell that the KT66 was overloading at high signal levels? How did you test that? What did you see?
When I short out R12 I see the 6L6GC plate dissipation goes up to 12W, which is still well within the limits of KT66, and is even within the limits of 6V6GT. The plate voltage and screen voltage are then both around 280V (which is fine).
7. The 12AU7 is running very current starved (Ip only 1.4mA), which means its rp must be somewhere up around 15k ohms. You put a plate load R8 of only 20k ohms, which is not even twice the 12AU7's rp at that operating point. That does give you low gain but it also brings high THD and possible problems driving the grid of the KT66 -- especially if it's wired in triode (higher input capacitance due to Miller effect). What led you to choose those operating points and parts values for the 12AU7?
8. If a design goal is to reduce the gain of the 12AU7 then why is its cathode resistor bypassed with a capacitor? Leaving that Rk (cathode bias resistor) without a bypass cap will introduce current feedback and both reduce the gain and lower distortion.
Last edited:
So if R6 is HP and not the amp's output Z, then that makes the output Z 470R pushing 300R. Not the right ratio, yes?
How did we figure the Zout is 470R?
Is that 1/gm of KT66?
...So I just gave it a quick check. Yes, in the sim I threw together the Zout is about 400 ohms. I would increase the gain of the voltage amp stage and wrap some parallel NFB around it to lower Zout to acceptable levels (which would also lower THD). You don't need super high input impedance. 50k would be good enough, since the sources are undoubtedly going to be either digital or a phono preamp.
Good catch!
Is that 1/gm of KT66?
...So I just gave it a quick check. Yes, in the sim I threw together the Zout is about 400 ohms. I would increase the gain of the voltage amp stage and wrap some parallel NFB around it to lower Zout to acceptable levels (which would also lower THD). You don't need super high input impedance. 50k would be good enough, since the sources are undoubtedly going to be either digital or a phono preamp.
Good catch!
Last edited:
It's the only significant resistance between the K and ground. I should say lowest R between K and ground.
Even with a CCS as the cathode load for the KT66 (6L6 in my sim), the Zout is only about 400 ohms.
I tested it by measuring a 1k sine wave output voltage into an open circuit (infinite load Z).
I got 3.4V peak out with 470mV pk signal in.
I decreased the load R until I got half the Vout (3.4V/2 = 1.7V).
That turned out to be about 400 ohms.
The Vout with a 300 ohm load was 1.4V.
I tested it by measuring a 1k sine wave output voltage into an open circuit (infinite load Z).
I got 3.4V peak out with 470mV pk signal in.
I decreased the load R until I got half the Vout (3.4V/2 = 1.7V).
That turned out to be about 400 ohms.
The Vout with a 300 ohm load was 1.4V.
hi @rongon out and about now and will respond more fully later but you can find the KT66 model I use at the beginning of this thread. I saw it was overloading by looking at the cathode voltage and current and multiplying them to get the wattage. The kt66 according to the spec sheet (also towards the beginning of this thread) has a max recommended anode wattage of 25W and absolute max of 30W, so over that you kill the tube.