[audioboozer]

It all began when I spent a ton of money on a ‘Great’ headphone amplifier. What was included in this supposedly hi-fi headphone amplifier was a single class-AB IC with 2 to 3 coupling capacitors, an OPamp with four transistors working as a buffer. Now its manual said that the headphone amplifier was capable to driving Sennheiser HD800 or equivalent headphone. That’s totally absurd! For example, HD800’s maximum long lasting input power is 500mW. To reach that output (of 500mW), your amplifier needed to deliver 35 Vpp (peak to peak voltage) for HD800! That means you needed an amplifier with at least ±18V working voltage (or a single DC voltage above 36V). Unfortunately, this is a common problem with high impedance headphones (Sennheiser HD800, or Beyerdynamic T1... etc.).

 

[AudioCats]

For high impedance phones? A "bulit-and-ready-to-use" Bottlehead Crack can be had for $300 from the "For Sale/Trade" forums here.

 

[audioboozer]

It can work both on high impedance and low impedance. $300 is a little be over budget. I am looking for something lower than $200. 

 

[audioboozer]

My dream headphone amplifier should consist of the most simplistic circuit as necessary (simple meaning using as few components as possible, and therefore creating less distortion), high out-put voltage and low output impedance running concurrently. Good harmonic characteristic is necessary for long-periods of listening and music enjoyment. The most important part is a wide frequency response that meets modern high resolution streaming music.
To meet these requirements, a class A, single-ended (SE) triode gain stage is a good start. Followed later by a PowerMOS source follower, which can deliver low output impedance to headphones.
The reason for choosing PowerMOS as an output device is because PowerMOS is a  “transconductance” device - a voltage controlled current device just like a tube.

 

[DutchGFX]

I designed and built my own tube amp from scratch this past year, and am more than happy to assist if you decide you want to do the same.

Mine was a 6SN7 input, 2A3 output, with a 10k reflected load on the tube for ideal power and low distortion into my Audeze LCD2.

I am clueless with SS design tho and admittedly limited in tube design as well but I can get the job done to some degree

 

[dsavitsk]

I am looking for something lower than $200. 

[COLOR=222222]My dream headphone amplifier should consist of the most simplistic circuit as necessary[/COLOR] [COLOR=222222](simple meaning using as few components as possible, and therefore creating less distortion), high out-put voltage and low output impedance running concurrently. Good harmonic characteristic is necessary for long-periods of listening and music enjoyment. The most important part is a wide frequency response that meets modern high resolution streaming music.[/COLOR]
[COLOR=222222]To meet these requirements, a class A, single-ended (SE) triode gain stage is a good start. Followed later by a PowerMOS source follower, which can deliver low output impedance to headphones.[/COLOR]
[COLOR=222222]The reason for choosing PowerMOS as an output device is because PowerMOS is a  “transconductance” device - a voltage controlled current device just like a tube.[/COLOR]

This sounds like the definition of Pete Millett's Starving Student, perhaps with an upgraded power supply. You can build it point to point, or there is a kit floating around :wink_face:

 

[audioboozer]

I would post my circuit within next few days. Then we can discuss about it together.

 

[audioboozer]

I would post my circuit within next few days. Then we can discuss about it together.

 

[audioboozer]

This circuit (shows one channel only) comes from many comprehensive tests. The input stage has a triode single-ended structure with constant current load, which provides good linearity and enough amplification gain. The constant current source is formed by a low noise, 1.7V red LED and Q5 transistor. A blue LED is series connected in order to light up the bottom of the tube. The output of this single-end stage is directly coupled to the Q6 transistor base, an emitter follower stage, then directly connected to the PowerMOS Q2. This provides enough driving capability to headphones. By using direct coupling, we not only reduced our cost, but also eliminate quality issues (for example: leakage current, ESL, ESR, etc.) from coupling capacitors.

 

[audioboozer]

Why not using output transformers?
Headphones have a wide range of impedance from 20 to 600 ohms. A low output impedance headphone amplifier is essential for driving low impedance headphones. A good amplifier should keep low impedance in all frequency range to ensure the same driving power is sent to all music spectrums. Even up to 100 KHz, it should keep the output impedance as low as possible. So, we need a high frequency response as well as low total harmonic distortion (THD). An output transformer simply won’t do the job!
Here is how the circuit designed. The power MOSFET (IRF610) has characteristics of a tube, which has high bandwidth with over 300 MHz and high input impedance.  Adding an emitter follower in front of the MOSFET can decrease the PowerMOS input capacitance loading-effects, and also increases its bandwidth. The IRF610 has input capacitance of 140 pF. By inserting this emitter follower, it can improve the bandwidth significantly. 
By the experiments, show that the bandwidth was tremendously increased with the implementation. With high bandwidth and low harmonic distortion, the amplifier was able to reproduce full range melodies and recreate music coming out from the headphones.

 

[tomb]

Sorry to say that you are woefully exaggerating the supposed disadvantages of output transformers. Maybe I'm not seeing your schematic correctly, but that one output-coupling capacitor in your circuit - C14 - is going to sound worse than most any output transformer you could use.

 

[audioboozer]

Choosing the Components

Adding a resistor on the emitter side of transistor (Q1) solves the current imbalance problem from the transistors. The two transistors (left and right channel) share one 13V Zener diode. This Zener stabilizes the voltage to 13V, which is necessary for the tube filament (12AU7 filament is 12.6V). Adding a parallel capacitor next to the Zener diode allows the capability of reducing breakdown noises from the Zener diode. 
Why use transistors with Zener diodes rather than regulator ICs? It is simply, better than using complex voltage regulator ICs! I also prefer using US made components like On-Semi, old Motorola, etc.
A high level of experience and expertise is vital to designing the perfect amplifier circuitry. A good measure of results may not truly reflect a great amplifier. However, good measuring results are still essential for developmental purposes. I believe an amplifier with good performance on top of excellent measured data is achievable.
Taking a look behind the scenes, what is behind the circuit structure of building the perfect headphone amplifier? It all comes down to selecting the right components!
For my headphone amplifier, I choose the best linearity transistors from On-Semi. Specifically, we decided to use from On-Semi, the MJE15030, the 2N5087, and even the Zener diode. In addition, we chose PowerMOS IRF610 from Visay, audio grade volume controller from ALPS, very low high-frequency impedance and long life capacitors from Panasonic, high performance capacitors from WIMA, and 6.3mm Neutrik headphone connector. 
Typically, it’s rare to see so many high-grade and high-cost components in commercial products.

 

[audioboozer]

Unfortunately, in my measurements, "the most any output transformer" is worse than capacitor except some very expensive output transformers.
 
Most transformers cannot have over 30 kHz frequency and phase response, but capacitor can easily pass 100 kHz with good linearity. Some very expensive transformers can reach 80 kHz at small signal level, but not linear when signal level changes.
 
This design has the flat frequency response up to 200 kHz with only -1dB, if memory tells me right there is no transformer can do so. I will share my figure later.
 
There are many "mythology" on audio related stuffs on internet. Please provide the scientific measurement data to prove your opinion. Otherwise, you are telling your own stories.

 

[tomb]

  Unfortunately, in my measurements, "the most any output transformer" is worse than capacitor except some very expensive output transformers.
 
Most transformers cannot have over 30 kHz frequency and phase response, but capacitor can easily pass 100 kHz with good linearity. Some very expensive transformers can reach 80 kHz at small signal level, but not linear when signal level changes.
 
This design has the flat frequency response up to 200 kHz with only -1dB, if memory tells me right there is no transformer can do so. I will share my figure later.
 
There are many "mythology" on audio related stuffs on internet. Please provide the scientific measurement data to prove your opinion. Otherwise, you are telling your own stories.

 
Well sure - if you want to design a headphone amplifier that responds all the way to 200 kHz, then your strategy is just fine.  If you want to design a headphone amplifier that responds best to what you can hear, then you're missing the boat.  As a device in the signal path, electrolytics are high in odd-order distortion.  Moreover, they "cloud" the entire output.  Even run-of-the-mill output transformers are more transparent than an electrolytic capacitor on the output.  The difference is not really subtle.  It's immediately obvious in direct comparisons. 
 
The problem is that you're over-designing when the result is that your C14 capacitor will dictate most of the sound quality of the entire circuit.  If you don't want output transformers, fine.  I think that's your loss to close off that possibility, but it's a design choice that's yours to make.  I just think if you're going on about the rest of the circuit, you would be better off designing some way to get rid of the circuit's biggest fault: C14.  Maybe investigate the use of a DC servo if you don't want an output transformer.  Right now, C14 is a band-aid over the entire circuit.
 
I'm not going to provide you with "scientific measurement data."  The internet is full of references about electrolytic capacitor distortion and lack of transparency.  Suffice to say there's ample evidence simply in the market for electrolytic capacitors: it's the reason they invented audio-quality electrolytic capacitors in the first place (Black Gates, Elna Cerafines, Elna RFS Silmics, Muse ES, Muse FG, Muse KZ, etc.,etc., etc.).  They improve upon the problem; they don't solve it.

 

[audioboozer]

Fine tuning!!
 
As I mentioned previously, headphones show different impedance under varies frequencies and/or voltages. Unless you have a very low impedance circuit, your amplifier may experience varies gains under different frequencies.
In order to solve this problem, I added some reverse feedback that can further lower the distortion, increase the bandwidth, and lower our output impedance. The R18 and R19 form the reverse feedback network and the overall gain is controlled about 10dB. Actual lab measurement shows that it can drive loading over 30 Ohms and none of the performance is lost.
 
Why did we choose to use just 10dB? This portion is where you can get the best fine tune over the volume. It will not run above the destination gain when tuning up and will not cause too little gain while tuning the volume down.
 
The trouble of using the dual triode tube, however, is that each half of the tube may not perfectly match the other. In other words, it may generate different gains from each half of the tube output. Question is, can you tolerate the 1dB volume difference between your left and right ear?
Our enhancement of our feedback network has successfully reduced the gain gap down to less than 0.1dB! This is important since the use of headphones placed right above the ears. Any minor gain differences can result in poor performance. This is why the designing a speaker amplifier versus building a headphone amplifier is different.