Suggestions for a very low distortion class A headphone buffer

[ionomolo]

I'm trying to build a class-A buffer for a transformer-based headphone amplifier. The idea is to have a very transparent buffer so the distortion of the transformer can be heard. The distortion of transformers is very nice from a theoretical point of view (it decays with frequency, which may allow to have a sweet midrange and clean voices), and has low intermodulation. It can also be increased by adding a small bias current.

After some playing in spice I've come out with the attached design. Distortion is excellent: 0.0001 % THD (almost all second harmonic) at 1 Vpk and 20KHz, which should drive my pair of AKG 701 to the unhealthy SPL of 102 dB's. It also has power to get to more than 110 dB's so there is headroom for peaks, which will have a slightly increased distortion. It should be noted that this is a huge improvement when compared to class-B operational amplifiers of the breed of the LM4562, which performs better at 1Vrms but distortion skyrockets at lower volumes which will constitute most of my listening, and it also contains less benign higher harmonics.

I'm having a couple of problems with this design. The first one is that I need a transistor for the output stage. The BC550B that I'm using in the simulations would blow up immediately if used with this current, but most power transistors are too powerful and too sluggish. I'm looking for something in the vein of the 2SC3505 but with slightly higher current capability (even if it comes at the price of lower breakdown voltage, certainly I do not intend to go near 300V). Any idea of how a power JFET would behave there? They seem very fast and don't have problems at 250 mA, but they are hard to find and poorly modeled in spice. Cascoding is another option but it comes at the price of decreased stability when compared to a single high-frequency transistor.

The second problem is not exactly a problem but an issue I'd like to improve on, and it is stability. The square wave response, while stable up to 500 pF loads, shows some nasty ringing that I'd rather not see. Obviously I can increase compensation until it goes away, but this would come at the expense of higher distortion. What would be really great is to make the amplifier faster, so the poles that are causing the ringing went upper in frequency were the gain is lower. I would appreciate very much hearing suggestions in this aspect.

Disclaimer: This is untested, design-stage schematic. While it is intended to work, it may need some modifications (the most notorious is the change of Q5 and the addition of a DC servo) and may present stability issues.

[Steve Eddy]

Originally Posted by ionomolo

I'm trying to build a class-A buffer for a transformer-based headphone amplifier. The idea is to have a very transparent buffer so the distortion of the transformer can be heard. The distortion of transformers is very nice from a theoretical point of view (it decays with frequency, which may allow to have a sweet midrange and clean voices), and has low intermodulation. It can also be increased by adding a small bias current.

You been reading Bill Whitlock?

se

[ionomolo]

Originally Posted by Steve Eddy

You been reading Bill Whitlock?

se

Yes. There are things about tube sound which I like, but I find the sound not clean enough. I've been wondering if using a transformer I could get the best of both worlds.

Do you have experience with transformers in audio? This dude is suspicious because of the interest in his company.

[Steve Eddy]

I think you'll be quite happy with the transformers. I like tubes too but found I like transformers better when it comes to voltage gain.

If you're not already familiar with them, check out CineMag (Audio Transformers By Cinemag Inc.). Specifically their CMMI-5C.

se

[ionomolo]

My idea is to use a 150:600 step up transformer, hence the 600 Ohm resistor in the source. I see both CineMag and Jensen do manufacture them. The CineMag has some nasty peaking at 150 KHz, but in the Jensen the distortion is quoted at -20 dBu, so I have no idea how they will perform at normal listening levels around 0 dBu.

CineMag:
http://cinemag.biz/mic_input/CMMI-2C.pdf

Jensen:
http://www.jensen-transformers.com/datashts/16a.pdf

Lundahl also makes transformers with higher turns ratio (and probably worse bandwith and distortion). If they costed the same as resistors I would buy one of each.

[Steve Eddy]

Originally Posted by ionomolo

My idea is to use a 150:600 step up transformer, hence the 600 Ohm resistor in the source.

Oh, ok. Don't know that I'd want to use a line output transformer as an input transformer. Hence my recommendation of the CMMI-5C.

I see both CineMag and Jensen do manufacture them.

That's because in a way, Jensen transformers are CineMag transformers.

Ed Reichenbach designed all the transformers for Deane Jensen back when Deane started Jensen Transformers. In fact, they were originally billed as "Jensen Transformers by Reichenbach Engineering." Deane split from Ed and basically took the engineering with him.

CineMag was started by the late Tom Reichenbach, Ed's son, and he has all the Reichenback engineering that was done for Deane.

The CineMag has some nasty peaking at 150 KHz...

What source material were you planning to feed it that has content at 150kHz?

Also note that the Jensen's using an RC snubber in parallel with 6.19k load resistor. So if that 150k resonant peak freaks you out, you can just use the same snubber on the CineMag.

...but in the Jensen the distortion is quoted at -20 dBu, so I have no idea how they will perform at normal listening levels around 0 dBu.

Well they both show THD versus frequency for fixed levels, so you can pretty much interpolate from there.

se

[ionomolo]

Originally Posted by Steve Eddy

Oh, ok. Don't know that I'd want to use a line output transformer as an input transformer. Hence my recommendation of the CMMI-5C.



That's because in a way, Jensen transformers are CineMag transformers.

Ed Reichenbach designed all the transformers for Deane Jensen back when Deane started Jensen Transformers. In fact, they were originally billed as "Jensen Transformers by Reichenbach Engineering." Deane split from Ed and basically took the engineering with him.

CineMag was started by the late Tom Reichenbach, Ed's son, and he has all the Reichenback engineering that was done for Deane.



What source material were you planning to feed it that has content at 150kHz?

Also note that the Jensen's using an RC snubber in parallel with 6.19k load resistor. So if that 150k resonant peak freaks you out, you can just use the same snubber on the CineMag.



Well they both show THD versus frequency for fixed levels, so you can pretty much interpolate from there.

se

In fact they are sold as mic-in transformers and not line out, pretty much like the one you linked me to, only they have less gain (which is OK since I never listen to my sources at full power. I guess 2x gain is already fine)

My concern for the 150KHz peaking is probably coming from solid-state design experience, which would indicate poor stability. But I will definitely add the snubber just to be sure that there is no digital high-frequency noise, where the amplifier is less stellar, getting augmented by the transformer and then demodulated by the amplifier.

Since you seem to be really into audio transformers, do you have any reason to prefer CineMag to Jensen?

By the way, I appreciate your comments about transformers and they are going to be helpful for my amplifier, but I would like to restate my questions about the solid-state output stage, since it is where I'm really stuck and I think I'm quite close to do significantly better than the most headphone buffers out there.

I reattach the schematic in PNG format for those who don't like to open PDF's.

[Steve Eddy]

Originally Posted by ionomolo

In fact they are sold as mic-in transformers and not line out, pretty much like the one you linked me to, only they have less gain (which is OK since I never listen to my sources at full power. I guess 2x gain is already fine)

Yeah, I know. But when you said "150:600," first thing that comes to my mind is a line output transformer wired 1:2.

My concern for the 150KHz peaking is probably coming from solid-state design experience, which would indicate poor stability.

Another nice thing about transformers. They don't oscillate.

Since you seem to be really into audio transformers, do you have any reason to prefer CineMag to Jensen?

Because I find them to be just as good, cost less, and when I call CineMag, I get the president of the company on the phone. Not some hired tech.

By the way, I appreciate your comments about transformers and they are going to be helpful for my amplifier, but I would like to restate my questions about the solid-state output stage, since it is where I'm really stuck and I think I'm quite close to do significantly better than the most headphone buffers out there.

I'm working on my own for a commercial amplifier so I'll leave that for others to respond to.

se

[volume knob]

Hi

If you're not apposed to I.C's then the buff634 in high bandwidth mode could be a simple solution

[bear]

Look at "the wire" thread here... it uses a bunch of ultra-super low distortion buffer chips in parallel as a power amp output stage, iirc. You could use just one.

Also, I would expect ringing into a capacitive load with an inductor at the output.
You don't need that if the circuit will not destroy itself into an ultrasonic virtual short circuit (very low Z) - which is merely a capacitive load...

Consider paralleled devices if you need speed and can not find speed in a single device?

_-_-bear