Surely, the point of building a simple tube headphone amplifier these days is because it is simple, it introduces simple distortions, is stable and sounds good. i.e. its virtue is its simplicity. Adding significantly more solid-state electronics defeats the purpose and is likely to introduce a whole host of other problems for you to deal with.

if the load is not 300ohms the push-pull tubes operate differently causing distortion when the signal rejoins.

What measurements have you taken to establish that this effect is significant and needs fixing?

If you think you need a more load tolerant amplifier; build a solid-state amplifier with a low output impedance, it will be much easier. As has already been mentioned, modern headphones are designed to be driven from low impedance sources not ones that change constantly.

 

Surely, the point of building a simple tube headphone amplifier these days is because it is simple, it introduces simple distortions, is stable and sounds good. i.e. its virtue is its simplicity. Adding significantly more solid-state electronics defeats the purpose and is likely to introduce a whole host of other problems for you to deal with.

Sounds more like you are hating on the complexity than anything else. I don't see the issue. I'm more interested in results than simplicity.

What measurements have you taken to establish that this effect is significant and needs fixing?

Me personally none, but I know people with my exactly same amplifier that experimented with changing the plate resistor value to more closely match their specific headphones and they have been impressed with the results, plus the tubecad guy says it's worth it, and the tubecad guy is a tube circuit legend.

As has already been mentioned, modern headphones are designed to be driven from low impedance sources not ones that change constantly.

Except the output impedance IS changing constantly due to a changing headphone impedance, the tubes operation point changes with headphone impedance. Hence my desire to stabilize it.

 

Except the impedance IS changing constantly due to a changing headphone impedance, the tubes operation point changes with headphone impedance. Hence my desire to stabilize it.

The impedance is different for each frequency, so the headphones represent a different load for each part of the spectrum that you are playing at the same time. You are not listnening to a single sinewave, so how do you choose WHICH of the infinitely many impedances is what determines your compensation value?

As you can see from the graph you posted, the impedance is 400 ohms or so from somwhere around 400Hz upwards, and around 600 near the resonance peak at 100Hz. I would choose such resistors that works well within these limits. Or use an amplifier that does not change its parameters significantly with load impedance.

But since you want to go with an extremely simple tube amp you cannot be surprised that the sound quality and introduced distortion will be very much determined by the combination of the amp and headphones, so you end up with fiddling around and choosing a value that you like the best. I don´t think any kind of compensation in the way you describe is physically possible.

 

Sounds more like you are hating on the complexity than anything else. I don't see the issue. I'm more interested in results than simplicity.

I'm not hating anything, I wrote: "Adding significantly more solid-state electronics defeats the purpose and is likely to introduce a whole host of other problems for you to deal with." Will this lead to good results?

Me personally none, but I know people with my exactly same amplifier that experimented with changing the plate resistor value to more closely match their specific headphones and they have been impressed with the results, plus the tubecad guy says it's worth it, and the tubecad guy is a tube circuit legend.

This is not the same as dynamically altering the operating point. If you don't know what the baseline performance of the amplifier is, how will you know if your new implementation has improved or worsened the performance?

Except the impedance IS changing constantly due to a changing headphone impedance, the tubes operation point changes with headphone impedance. Hence my desire to stabilize it.

You're the expert.

 

If you want the current proportional to the voltage at all impedances, connect the signal to the grid of a pentode and add a cathode resistor. That creates a constant current generator. Connect the headphones in the plate circuit. It will totally screw the intended design of the headphones, but that doesn't seem to be your criteria. When you find that the headphones are more efficient at high frequencies to compensate for the impedance change, you will know why good engineers spent thousands of hours creating the headphones.

In post #13, sailorjoe described the usual method to achieve a mostly constant current response 60 years ago. When the impedance of the load is insignificant (less than 10%) compared to the resistance in series with it, changes in impedance become proportionally less significant.

 

The impedance is different for each frequency, so the headphones represent a different load for each part of the spectrum that you are playing at the same time. You are not listnening to a single sinewave, so how do you choose WHICH of the infinitely many impedances is what determines your compensation value?

Yeah, that's why I said it might be a good idea to measure the frequencies and their amplitudes to create an average and use that average to create a reference voltage. The tube can't have multiple simultaneous operating points therefore it must change based on an average.

But since you want to go with an extremely simple tube amp you cannot be surprised that the sound quality and introduced distortion will be very much determined by the combination of the amp and headphones

For the record, the amplifier sounds unimaginably good, which is why I'm obsessed with improving it in any way I can

I don´t think any kind of compensation in the way you describe is physically possible.

Is there something wrong with my sampling idea?

I'm not hating anything, I wrote: "Adding significantly more solid-state electronics defeats the purpose and is likely to introduce a whole host of other problems for you to deal with." Will this lead to good results?

Who knows, I'm hoping to find out. You still haven't told me what these host of other issues are.

This is not the same as dynamically altering the operating point. If you don't know what the baseline performance of the amplifier is, how will you know if your new implementation has improved or worsened the performance?

I'll test the frequency response at the output based on varying frequencies with my headphones plugged in. Also I'll test the operating point of both tubes to make sure they are running at the same point at all frequency averages.

You're the expert.

If you think I am wrong you should tell me why.

If you want the current proportional to the voltage at all impedances, connect the signal to the grid of a pentode and add a cathode resistor. That creates a constant current generator. Connect the headphones in the plate circuit. It will totally screw the intended design of the headphones, but that doesn't seem to be your criteria.

I don't want current proportional to the voltage, I want the tubes to function at the same operating point, I fail to understand why what I'm doing is going against the intended design. As is, the operating point between the two push pull tubes shift with a changing output impedance, hence the split signal is working through tubes that are out of sync with each other. All I want to do is ensure they are always working at the same operating point which will provide stable operation as if the headphone impedance was static at all frequencies like a pair of planar headphones.
For example, I want the top and bottom triode to stay in their load line "sweet spot" throughout their operation which for my specific tubes happen to be 70volts at 75 miliamps. The bottom tube stays at a stable operating point but the top tube changes with load impedance because its plate resistor value is balanced against the bottom tube based on the intended load impedance.
So If I have a plate resistor value set to a 300 ohm load impedance and the load is 300 ohms then the operating point will stay at 70v 75ma, but if it shifts from 300 ohms it will end up at a different current/voltage while the bottom tube maintains its 70v 75ma.
Surely the output impedance must change with differing operating points.

 

I fail to understand why what I'm doing is going against the intended design.

That is obvious. Furthermore, refusing to believe the advice given tends to make people not want to continue in their futile efforts to educate you.

 

You haven't given me any specifics behind the advice. Read my above post and explain to me why it is wrong. I'd be glad to be wrong because it saves me time and money, for me this is a chore not a hobby, but I'm not going to accept "just because".

You seem to think I want constant current, I can care less about constant current. I want to match the power output of both of the WCF tubes to achieve a perfect push-pull. I don't see how this is doing anything but stabilizing the output impedance of the amplifier and reducing the distortion.

 

...but I know people with my exactly same amplifier that experimented with changing the plate resistor value to more closely match their specific headphones and they have been impressed with the results, plus the tubecad guy says it's worth it, and the tubecad guy is a tube circuit legend.

Can we assume that those that changed the plate resistor set it for a fixed value, but didn't change it in real time? I can see how making the adjustment to match specific headphones or speakers would work well. Has anyone before figured out how to vary plate resistance in real time to match the music? (I did read the article on power adjustment in near real time to match the volume of the music, by the way.)

Except the output impedance IS changing constantly due to a changing headphone impedance, the tubes operation point changes with headphone impedance. Hence my desire to stabilize it.

So here's the dilemma. The headphone impedance is fixed, but it is clearly a function of frequency, so the only way to match the operating point to the impedance is to reduce the signal to a single frequency. To do that, we'd have to calculate the instantaneous frequency (https://www.math.ucdavis.edu/~saito/data/sonar/boashash1.pdf) of the signal, and use that to match the operating point. To do that, we'd need to sample the signal and make an adjustment at something around 100KHz. A 100 KHz sample is 10 microseconds long, And you can see from the article in the link that the math to do all this is fairly exotic. If we can find or develop the sampling and instantaneous frequency circuits, then we could figure out how to adjust the operating point. The last time I saw this being done was a project for DARPA in New Jersey, using millions of dollars of research.
The next option I can think of is to break the audio band into subsets which correspond to sections of the impedance curve of the headphones where the variation is relatively small. We measure the total energy of each band in real time, and select the band with the largest energy. The selected band is used to control the operating point. The system would have the tendency of responding to the loudest frequency group, in general, which might match well with music. It might not, though, if low heavy bass drum hits swamp the melody line, but perhaps it's worth a try. In fact, it might be possible to set the operating point to match the low end of the headphones, and then to run mid range signals through and see if they distort.

This project is pushing the edge of what's been done before with audio tubes, so there are no easy answers to the problems you raise. In fact, there may not be any complicated answers either, since it hasn't been done. Who knows, there could be an interesting breakthrough here, eventually, but I'd expect to hit some failures before finding it. I don't mind trying to help you find options; you seem willing to do the work. Just be aware, we are beyond standard practice as described by me and others, and into the realm of the unknown......and the expensive.

 

https://www.math.ucdavis.edu/~saito/data/sonar/boashash1.pdf
That math makes me want to kill myself
I don't know anything about digital circuitry but surely it shouldn't be too hard to create a logic circuit that takes the voltage amplitudes and frequencies during a time frame and averages them, maybe.

Just be aware, we are beyond standard practice as described by me and others, and into the realm of the unknown......and the expensiv

Ah but the unknown is where we find hidden treasures, I hope.

Here is an esteemed article on the very subject I am trying to correct, it's good stuff. My eyes start glazing over during the math so I haven't finished reading it yet
https://www.cavalliaudio.com/diy/docs/WCFOptimization.pdf
but it confirms my point.

 

Well, I had to learn all that math, and much more, to get my degree. So scary or not, it's how this stuff really works. And fortunately there is, in fact, a circuit that can analyze the signal and tell you what's happening in the frequency domain. It's called a spectrum analyzer. It costs $1000-$100,000 dollars, but it's too slow to work in this application. So we turn to getting the instantaneous frequency instead. That can be done by using a VCO in a PLL (all analog). Then we have to use that information through some sort of strange function generator that matches headphone impedance. Then that runs through a control circuit to manage the plate resistance.
Personally, I prefer the idea of the multiple band analyzer. It's a lot like an equalizer, and you only need about six bands to make it work. They don't have to be equally spaced. And it's also all analog. By the way, all of these suggestions are things outside of the primary signal path. We tap the signal as it comes in, without distortion, process it on the side, but we don't insert anything into the signal path. I assume that meets your expectations.
I'll read the article tomorrow. Thanks for sharing it.

 

I don't know anything about digital circuitry but surely it shouldn't be too hard to create a logic circuit that takes the voltage amplitudes and frequencies during a time frame and averages them, maybe.

But why, when you already have the complete characteristic of the headphones you have? Do you plan to use the amp with lots of different headphones? I see no point in measuring the impedance on the fly and somehow adjusting the amplifier (which won´t be very easy either). Like I said, choose a value that matches the average of the impedance measurement you already have and be done with it.

Or maybe use a switch to choose between a few different settings and decide what you like best.

 

That can be done by using a VCO in a PLL (all analog). Then we have to use that information through some sort of strange function generator that matches headphone impedance. Then that runs through a control circuit to manage the plate resistance.
Personally, I prefer the idea of the multiple band analyzer. It's a lot like an equalizer, and you only need about six bands to make it work. They don't have to be equally spaced.

I need to do research on what you just said

But why, when you already have the complete characteristic of the headphones you have? Do you plan to use the amp with lots of different headphones? I see no point in measuring the impedance on the fly and somehow adjusting the amplifier (which won´t be very easy either). Like I said, choose a value that matches the average of the impedance measurement you already have and be done with it.

I'm currently only tuning it for the HD800s. The reason I need to measure frequency on the fly is because the output stage's push pull operation loses its synchronization when the impedance changes. The tube can only have 1 operating point at a time, so it must change it's operating point based on an average of impedances on the load based on what frequencies are running through it at any one moment.

Therefore, in order to maintain synchronized push pull the operating point of one of the tubes must change to match the other when the load impedance changes.

I certainly could just use a resistor based on an average value but that would be the easy way out. What I'm trying to do is about optimization.
I care more about the bang than the buck. I think that applies much in audio, people will do anything to achieve that perfect sound. Who knows how much of an improvement this will make but any improvement is a good improvement.

I'd like to point out that my amplifier as is sounds so good I could not have imagined it being possible without hearing it myself. It sounds 99% like real life on a quality recording, I can actually believe I'm "there" if I close my eyes and my brain gets confused sometimes because it sounds like a person is standing in front of me.
I'm not trying to make a bad amplifier sound good, I'm trying to make a great amplifier sound better.

 

But why, when you already have the complete characteristic of the headphones you have? Do you plan to use the amp with lots of different headphones? I see no point in measuring the impedance on the fly and somehow adjusting the amplifier (which won´t be very easy either). Like I said, choose a value that matches the average of the impedance measurement you already have and be done with it.

Or maybe use a switch to choose between a few different settings and decide what you like best.

Kubeek, I believe coinmaster is trying to push the state of the art in his amplifier. Whether it's possible or not is an open question. But hey, if he doesn't try, we'll never know. So why don't we help him take his amplifier as far as he can? No cost to us.

 

So the goal here is to mesure the instanteneous voltage and current, divide the two to get impedance, average the resulting impedance and then somehow feed this to the amp?

 

Something like that yeah, except there can be multiple frequencies with multiple voltage levels, so I'm not sure how that would work.
If I have a signal at 600 ohm impedance and another simultaneous signal at 300 ohm impedance, does the current flow average somewhere in between?
We could measure voltage and current, or we can just measure frequency and cross reference it with the impedance curve of my headphones which is probably the better solution.
I'm still figuring out the details of salorjoe's proposal.

 

The goal in selecting the operating point and component values for the WCF OTL stage is to achieve push-pull balance in the amplifying and regulating triodes. Balanced pushpull operation will maximize the power and minimize the distortion

The calculations in the current balance section above together with Eq. 28 make it clear that there is no single value of Ra that will balance the WCF for a range of loads and for both small and large signal regimes. Any choice of Ra is, therefore, a compromise to get the best possible behavior for these two separate constraints

Not for long

 

coinmaster, one option is to find headphones that have a flat impedance. Then match the plate resistor to that one impedance. Is that possible?

Do you want to pursue the instantaneous frequency solution, or the band filtering solution? Trust me, the band filtering solution is a lot easier to pursue. Neither solution can be guaranteed to work, but the band solution is far less work, and there are some relatively easy tests we can make to see if its worth pursuing to completion.

 

I doubt any headphones will have a flat impedance. Invariably all will have a peak somewhere around 100hz because of resonance in the driver mechanics, and all will roll off in the high end because of the inductance of the coil, plus additional peaks or vallyes in places where the designer tweaked the performance.
Maybe electrostatic might be better, but that is a whole different can of worms.

 

coinmaster, one option is to find headphones that have a flat impedance. Then match the plate resistor to that one impedance. Is that possible?

Planar headphones have a flat impedance from what I hear. But the HD800s are legend and when driven by the right amplifier they are the most amazing sounding headphones I've ever heard, and I've heard almost all of the high end options. I don't think the issue I'm trying to solve is significant enough to warrant a change in headphones

Do you want to pursue the instantaneous frequency solution, or the band filtering solution? Trust me, the band filtering solution is a lot easier to pursue. Neither solution can be guaranteed to work, but the band solution is far less work, and there are some relatively easy tests we can make to see if its worth pursuing to completion.

Yeah I'm still trying to figure out the details behind those plans. You want to use a vco to go into a function generator that goes into the voltage control which I'm kind of confused on, is the vco just used as some kind of buffer? How would the function generator help?
Also the band analyzer which is basically the same thing as a spectrum analyzer isn't it? Sorry my experience in electronics is months old.