• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Tube Headphone Amp Design and Help With Inverse Pre-Distortion Circuit

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi all!

I decided to try designing a tube headphone amplifier using the "inverse complementary distortion cancellation" technique that's described in the TubeCAD journal: http://www.tubecad.com/articles_2001/Inv_Dist_Cancellation/CompInvDisAmp.pdf (It also gets a mention in Morgan Jones' Valve Amplifiers book on page 180).

The gain stage uses both triodes in a 6922 tube in a common cathode -> cathode follower configuration with SS current sources and a DC servo to ensure that both triodes see the same bias conditions. The output stage uses both triodes in a 6AS7 tube in a white cathode follower configuration with another DC servo to eliminate the output coupling capacitor. There is no global feedback.

Overall, I think I'm getting pretty decent performance out of this amplifier in the simulations (especially when I adjust the plate load on the output triode, R11, to be optimized for the load impedance it's driving -- I might add a switch to select between low impedance and high impedance headphones when I build it), but the distortion cancellation technique doesn't seem to be doing anything useful. I get roughly the same performance (actually slightly better) when I throw out the cathode follower stage where the distortion cancellation is supposed to take place!

Does anyone with experience in this type of circuit see any problems with what I've done? I've tried adjusting U3's bias resistors (R6, R7, R14) to see if it's different AC load conditions between U1 and U2 that are the problem, but it doesn't seem to affect the results much.

Also, I'd be very grateful if I could get some comments on the rest of the design, particularly the output stage. I've tried playing around with different bias points and using a parallel cathode follower instead the WCF topology, but what I have right now seemed to work best in the simulations.

This is my first time designing with tubes from scratch :)
 

Attachments

  • headphone_tube_amp_6922_6as7_wcf_2servo_rev5_schem.png
    headphone_tube_amp_6922_6as7_wcf_2servo_rev5_schem.png
    117.9 KB · Views: 204
I don't think that Tubecad article has this right. A cathode follower is fully neg. feedback degenerated, so does not generate useful complementary (2nd harmonic) distortion for the grounded cathode stage. Especially with a CCS load on the follower.

These distortion cancellation schemes generally only cancel 2nd harmonic, even when they do work. And they require the 1st stage to distort so badly to get enough 2nd H, that it ends up making lots of 3rd and higher odd harmonics too.
By the time the 1st stage's 2nd harmonic matches the 2nd stage, the combined characteristics look like an S shape curve.

For a real comprehensive distortion cancellation scheme for all harmonics, the two stages would have to track current and voltage wise to provide equal (but opposing) distortions. Only way to do that is by using the same tube element right in a neg. Fdbk path. And then you are lucky if they track well enough for more than just a few low harmonics. Since simple conventional N Fdbk can clean up the lower harmonic distortions so easily anyway, it seems somewhat futile.
 
Administrator
Joined 2004
Paid Member
I think both of the previous posters hit the nail squarely on the head. I'll add the coffin.. In the event of a servo failure on the output stage you are likely to get a pretty big DC offset, headphone killing, and potentially injurious to your hearing and could be a shock hazard as well. DC coupling isn't such a good idea in a headphone amp.

With CCS loads on the VAS and a well designed WCF (no tricks) distortion will be so low you won't care.
 
Thanks for the quick responses guys.

I don't think that Tubecad article has this right. A cathode follower is fully neg. feedback degenerated, so does not generate useful complementary (2nd harmonic) distortion for the grounded cathode stage. Especially with a CCS load on the follower.

These distortion cancellation schemes generally only cancel 2nd harmonic, even when they do work. And they require the 1st stage to distort so badly to get enough 2nd H, that it ends up making lots of 3rd and higher odd harmonics too.
By the time the 1st stage's 2nd harmonic matches the 2nd stage, the combined characteristics look like an S shape curve.

For a real comprehensive distortion cancellation scheme for all harmonics, the two stages would have to track current and voltage wise to provide equal (but opposing) distortions. Only way to do that is by using the same tube element right in a neg. Fdbk path. And then you are lucky if they track well enough for more than just a few low harmonics. Since simple conventional N Fdbk can clean up the lower harmonic distortions so easily anyway, it seems somewhat futile.

These are great points, and I think you've convinced me to do away with this idea and just use the second triode in the other channel to save a tube. I tried disconnecting the cathode follower from the second stage to make it easier to vary its AC load (I put a resistor to ground after C1 instead), and I did find that I was able to achieve some amount of distortion cancellation in H2 with the right conditions, but it wasn't much, and the result was that now H3 was a lot more significant. The ideal CCS in the simulator has infinite output impedance, so there is basically no distortion from the cathode follower when it's unloaded, but things are different when there is an AC load.

You've got tubes with beautiful curves, and you're ok to use SS components to assist them. Why do you want to use elaborate distortion canceling schemes, when you can just set up your tubes not to make much distortion in the first place?

Yes, you're right, I picked the 6922 tube for it's low distortion already. I was feeling curious to see how well I could do :)

I think both of the previous posters hit the nail squarely on the head. I'll add the coffin.. In the event of a servo failure on the output stage you are likely to get a pretty big DC offset, headphone killing, and potentially injurious to your hearing and could be a shock hazard as well. DC coupling isn't such a good idea in a headphone amp.

With CCS loads on the VAS and a well designed WCF (no tricks) distortion will be so low you won't care.

I agree that there's a certain elegance to simplifying the circuit as much as possible and just using an AC coupling capacitor, but I think that the DC servo approach is worth it in this case. I'll admit that I don't have a pair of golden ears or lots of listening experience, but I think that eliminating the distortion from this capacitor (as well as saving on its cost) will be noticeable, if anything from the various threads on HeadFi about listening to different coupling capacitors is to be believed... If we switched to a single ground-referenced supply, then this coupling capacitor would have the highest value, greatest DC bias, greatest AC current, and be approximately tied for the greatest AC voltage swing of all of the capacitors in the circuit.

I do plan to implement an output protection circuit with comparators and a relay just in case, but even in all of the worst-case scenarios I can imagine, nothing terrible happens if the DC servo fails. The extreme failure corners I can imagine are Q1 CE fails short, Q1 CE fails open, Q1 C is shorted to +24V, and Q1 C is shorted to -24V. I simulated all of these with and without an input signal and at the two headphone impedance corners I've been working with (30 ohms and 300 ohms). The worst-case scenarios are that DC offsets of about 1.4V and 5.3V develop at the output with 30 ohms and 300 ohm loads, respectively. R6, R7, and R14 are sized so that the DC servo only has the ability to change the grid bias by about 10 volts in either direction about -20V. These offsets (along with any additional small offsets that occur due to the high THD in these scenarios) result in a few tens of mW of DC power being dissipated in the headphones for a few seconds before a comparator would catch the error by sensing the output of opamp U5. Even if we assume that the comparator circuit is broken or not installed, I wouldn't be too concerned about this. In the former case I would also OR in a second comparator sensing the output node directly to quickly catch any excessive voltage swings.

Overall, my philosophy here is that I would rather put in a simple circuit made of $10 of SS components than spend time and money rolling that output capacitor. I think if it works like it's supposed to, then it can only be an improvement.

Do you think I've missed anything here?
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.