Hello all. I am working on an OTA, OTL, OCL, OCD : P headphone amp for a friend. He loves his HD800s and the impedance never drops below 300 ohms, so I have the following circuit idea.
We have a couple problems with this circuit. 1: I would need to interface the incoming ac signal with a circuit that is sitting at ~ -155v 2: The input capacitance of the 4P1Ls would be pretty high and require a fairly beefy driver.
Both requirements can be solved separately, however, is there a way were we can kill two birds with 1 stone? Short answer? Maybe. I'm not really sure and I could use some help.
Here is my test circuit. The idea is to create a cathode follower where the output sits at ~ -155v. The signal is fed to G2 instead of G1. This allows G2 to remain at ground potential whilst G1 and the cathode can be shifted down towards the negative supply rail. Finally, I am trying to incorporate schade feedback in order to have triode like performance at the output.
In this circuit the cathode does sit at the correct potential! Yay. Level shifting achieved!
Only 1 small problem. I hardly get any output signal. Where I would expect to get nearly 1v out (0.95v ish), I only get 6mv out 😢
I am hoping some of you might be able to enlighten me on what I might be missing here.
We have a couple problems with this circuit. 1: I would need to interface the incoming ac signal with a circuit that is sitting at ~ -155v 2: The input capacitance of the 4P1Ls would be pretty high and require a fairly beefy driver.
Both requirements can be solved separately, however, is there a way were we can kill two birds with 1 stone? Short answer? Maybe. I'm not really sure and I could use some help.
Here is my test circuit. The idea is to create a cathode follower where the output sits at ~ -155v. The signal is fed to G2 instead of G1. This allows G2 to remain at ground potential whilst G1 and the cathode can be shifted down towards the negative supply rail. Finally, I am trying to incorporate schade feedback in order to have triode like performance at the output.
In this circuit the cathode does sit at the correct potential! Yay. Level shifting achieved!
Only 1 small problem. I hardly get any output signal. Where I would expect to get nearly 1v out (0.95v ish), I only get 6mv out 😢
I am hoping some of you might be able to enlighten me on what I might be missing here.
The screen grid has extremely low mu, that's why the gain is so low. Is this one of those situations where you're fixated on avoiding a simple coupling capacitor?
Not to mention the danger to the headphones.
At least the audio cults don't have the same consequences as some modern political cults.
All good fortune,
Chris
At least the audio cults don't have the same consequences as some modern political cults.
All good fortune,
Chris
I love how quirky the whole thing is, but I agree with others that the added complexity and safety might be taking one step forward and two back. Having at least three power supplies means that you're introducing at least three caps to the current loops to save one. With DC coupling, it's also going to be important to use some circuitry to set the DC conditions as the tubes warm up and as they age-- gyrators/chokes, self bias, a servo, certainly a muting relay.
You mention you'll be driving HD800s. The 4P1L has a mu of 8, so if you plug a DAC that puts you 2VRMS, that's 16VRMS into the load. Are you sure that you need a driver with gain and not just a buffer?
Rp is also fairly high: 8 tubes per channel will get you an output impedance of 150 ohms while eating a gargantuan 5.6A of heater current per channel. Efficiency may not be a design goal, but I would experiment with driving the HD800 through an amp with a low output impedance and a 150 ohm series resistor to see if this gets you the sound you're after. My experience is that some output impedance does sound good, but more than 50 ohms will be detrimental to the sound.
You mention you'll be driving HD800s. The 4P1L has a mu of 8, so if you plug a DAC that puts you 2VRMS, that's 16VRMS into the load. Are you sure that you need a driver with gain and not just a buffer?
Rp is also fairly high: 8 tubes per channel will get you an output impedance of 150 ohms while eating a gargantuan 5.6A of heater current per channel. Efficiency may not be a design goal, but I would experiment with driving the HD800 through an amp with a low output impedance and a 150 ohm series resistor to see if this gets you the sound you're after. My experience is that some output impedance does sound good, but more than 50 ohms will be detrimental to the sound.
To address everyone.
Yes, I do what I can to avoid capacitors. The man I am building this amp for would inevitably spend wayyyyyy too much money and time swapping capacitors around. It hurts my heart to watch it happen, so I try to avoid it wherever possible.
Yes, this will have some sort of DC servo in the long run. I never bother adding control circuitry to a schematic until I have the foundation done first.
Yes, some gain is required. The hd800s are weird and actually do want a ton of voltage. We tested them by listening to them as loud as we felt comfortable (emphasis on comfortable and less on loud) and saw peaks of 12V rms. This is the upper threshold, so if I can output 12vrms, we should never have to worry about headroom or clipping.
The 4P1L does have a higher plate impedance than what I would like to see. Absolutely true. Thankfully the tube is so linear that even with a 300ohm load, the load line against 8 tubes isn't terrible looking. To have a meaningful reduction in impedance, I would need to use more sometin more expensive such as some 2a3/300b-ish tube and I would still need to have 8 in parallel. Only something like 6SP45 (EL509) or the 6P44S would reduce my number of tubes while still still having nice evenly spaced grid lines, but you run into a different problem.....
.... The heater power. You are correct that 8 4P1Ls would need 8 x 650ma = 5.2amps of current to run. But you missed the part where it only needs 2.1 volts. So the total wattage to run all 8 filaments is only ~11 watts. 1 single EL509 needs 2.5 amps at 6.3 volts which is 15.75 watts 😱. So 8 4P1Ls in parallel per channel really isn't all that crazy. I can actually do some level of filament bias and still have less wattage into the heater circuit compared to a few EL509s. The only gotcha there is.....
.....Miller capacitance. Honestly, out of all the insurmountable problems I face, this is the one that concerns me the most. Bartola writes this about the 4P1L capacitance.
"Capacitance between anode and first grid for 4P1L is 0.1 pF. Capacitance between screen grid and control grid is about 1 pF. It has a 10 pF Miller capacitance which is not high value, and for 20 KHz it is slight less than 1MΩ impedance. Any driver with 10 mA idle current will make it happy."
So lets say that the input impedance for this tube at 20K is 1Mohm which means that if we need to consider 40Khz as our target frequency response in order to insure that our response at 20khz is flat without any phase shift or other nasty nonsense. This gives us an impedance of 400Kohms. Divide by 8 for our 8 4P1Ls and you get an input impedance of 50Kohms. Not great, not the worst either. Now we do get a little bit of a break here since we aren't trying to swing massive amounts of voltage. My goal is to output 12Vrms at 40Khz, and if we plug that into the formula I= 2pi * F * V * C I get 0.3ma peak. This seems so comically low, that I feel like my math is off. If Im not wrong, then hell yeah, I will use an input transformer and call it a day. But I have a feeling I am going to need to drive this thing with at least 5ma.
Lastly, the output impedance vs the load impedance is somewhat of a concern, but if I find the sound quality is lacking, I can tackle that problem later.
Yes, I do what I can to avoid capacitors. The man I am building this amp for would inevitably spend wayyyyyy too much money and time swapping capacitors around. It hurts my heart to watch it happen, so I try to avoid it wherever possible.
Yes, this will have some sort of DC servo in the long run. I never bother adding control circuitry to a schematic until I have the foundation done first.
Yes, some gain is required. The hd800s are weird and actually do want a ton of voltage. We tested them by listening to them as loud as we felt comfortable (emphasis on comfortable and less on loud) and saw peaks of 12V rms. This is the upper threshold, so if I can output 12vrms, we should never have to worry about headroom or clipping.
The 4P1L does have a higher plate impedance than what I would like to see. Absolutely true. Thankfully the tube is so linear that even with a 300ohm load, the load line against 8 tubes isn't terrible looking. To have a meaningful reduction in impedance, I would need to use more sometin more expensive such as some 2a3/300b-ish tube and I would still need to have 8 in parallel. Only something like 6SP45 (EL509) or the 6P44S would reduce my number of tubes while still still having nice evenly spaced grid lines, but you run into a different problem.....
.... The heater power. You are correct that 8 4P1Ls would need 8 x 650ma = 5.2amps of current to run. But you missed the part where it only needs 2.1 volts. So the total wattage to run all 8 filaments is only ~11 watts. 1 single EL509 needs 2.5 amps at 6.3 volts which is 15.75 watts 😱. So 8 4P1Ls in parallel per channel really isn't all that crazy. I can actually do some level of filament bias and still have less wattage into the heater circuit compared to a few EL509s. The only gotcha there is.....
.....Miller capacitance. Honestly, out of all the insurmountable problems I face, this is the one that concerns me the most. Bartola writes this about the 4P1L capacitance.
"Capacitance between anode and first grid for 4P1L is 0.1 pF. Capacitance between screen grid and control grid is about 1 pF. It has a 10 pF Miller capacitance which is not high value, and for 20 KHz it is slight less than 1MΩ impedance. Any driver with 10 mA idle current will make it happy."
So lets say that the input impedance for this tube at 20K is 1Mohm which means that if we need to consider 40Khz as our target frequency response in order to insure that our response at 20khz is flat without any phase shift or other nasty nonsense. This gives us an impedance of 400Kohms. Divide by 8 for our 8 4P1Ls and you get an input impedance of 50Kohms. Not great, not the worst either. Now we do get a little bit of a break here since we aren't trying to swing massive amounts of voltage. My goal is to output 12Vrms at 40Khz, and if we plug that into the formula I= 2pi * F * V * C I get 0.3ma peak. This seems so comically low, that I feel like my math is off. If Im not wrong, then hell yeah, I will use an input transformer and call it a day. But I have a feeling I am going to need to drive this thing with at least 5ma.
Lastly, the output impedance vs the load impedance is somewhat of a concern, but if I find the sound quality is lacking, I can tackle that problem later.
...''Lastly, the output impedance vs the load impedance is somewhat of a concern, but if I find the sound quality is lacking, I can tackle that problem later.''
Isn't that the point of the build? 300 ohm phones do best with 1/10, like most other components. What is your target output Z?
Isn't that the point of the build? 300 ohm phones do best with 1/10, like most other components. What is your target output Z?
Sorry for not replying, I have been sick.
Yes, a lower output impedance would theoretically be better, but dampening factor of 2 isn't the end of the world. Granted I am kind of banking on the idea that the impedance of the headphone is ~500 ohms at 30hz.
Obviously I could make a bog standard OTL amp with one of the ultra low impedance tubes. But where is the fun in that? I have heard probably 3 or 4 dozen different OTL amps and none of them really work for me. They all sound really dry and clinical minus the greek amp and a couple of the atmosphere designs. I have never seen or heard an anode follower OTL, so I am curious to try it.
Now. If the 4P1L anode follower doesn't work, I will follow up with Unset to drop the output impedance a bit more. If THAT doesn't work, I will keep adding 4P1Ls until we find a sweet spot. If we make it to 16 tubes without any success, well then I have a large stock pile of 4P1Ls. Not a big concern to me 😉
I will be comparing this amp to another amp I am making that uses 4x6P45S tubes in parallel.
If neither of those sound good, I am abandoning the whole thing because the power requirements will indeed become stupid at some point.
Yes, a lower output impedance would theoretically be better, but dampening factor of 2 isn't the end of the world. Granted I am kind of banking on the idea that the impedance of the headphone is ~500 ohms at 30hz.
Obviously I could make a bog standard OTL amp with one of the ultra low impedance tubes. But where is the fun in that? I have heard probably 3 or 4 dozen different OTL amps and none of them really work for me. They all sound really dry and clinical minus the greek amp and a couple of the atmosphere designs. I have never seen or heard an anode follower OTL, so I am curious to try it.
Now. If the 4P1L anode follower doesn't work, I will follow up with Unset to drop the output impedance a bit more. If THAT doesn't work, I will keep adding 4P1Ls until we find a sweet spot. If we make it to 16 tubes without any success, well then I have a large stock pile of 4P1Ls. Not a big concern to me 😉
I will be comparing this amp to another amp I am making that uses 4x6P45S tubes in parallel.
If neither of those sound good, I am abandoning the whole thing because the power requirements will indeed become stupid at some point.
It really shouldn't necessary to go that far to drive 300R headphones. Plenty of people build headphones amps with a single 6DJ8 driving such phones e.g. White cathode follower or SRPP. If a single tube can do it, you really don't need 16 tubes 😵
Your friend has an expensive taste in headphones. Does he not have something to plug them into now that is up to the task, something that you believe you can build that would be a definite improvement? If he doesn't have any tube equipment now, you could kill 2 birds and build a SE amp with an impedance matcher on the outputs. Like you said, the cost of your project could get stupid. It's not going to be much different than a conventional amp which is already proven. You can't really reinvent tube technology. Just make it more complex to get to the same point. You can get to New York from Philadelphia by going west, too.
Right. And yeah, obviously it would be nice to use fewer tubes. Im simply stating that going up to 16 tubes isn't off the table even if it's just a test.
Why would any of this be "expensive"? 4P1Ls are ~4-6 bucks a tube. Absolute worst case we are looking at around 400 bucks for the whole amp, but realistically we are looking at ~150ish hopefully.
And he has plenty of amps for these headphones. Im literally building this as a gift since his birthday and christmas are around the same date. But this is in addition to his 45, 71a, 6l6, 300b, 300b push pull, hybrid, SS, esl, and circlotron amps. We just like collecting different topologies.