Headphone Amp Redux

A quick google found the info:

Scroll down to the 25w music amplifier section.

They used 6082 tubes, which were 26.5v heater versions of the 6080, so that they could run all the heaters in series across the line and they only used 3, where my memory suggested more. Whatever, the frequency response is pretty amazing for a tube amp and the harmonic distortion is not bad either. Could be improved by using more tubes in parallel, one suspects

If you are in to tube amps, a good project for those cold winter evenings, perhaps ?...

Regards,

Chris

What happens if you remove them (since most amplifiers do without anyway

- especially high powered amps into low ohms where it is even more critical)?

I'd like to see something similar for microconroller instruction set design -- a "nearly useless instruction set processor," where the instruction set is designed to be as nearly useless as possible, but with each instruction not entirely useless. NOP is far, far too useful by itself, so it would probably be eliminated in a qualifying design, requiring instead some stream of instructions to achieve it.

Imagine spreading out all of the usual "useful" instructions at verticies of a vast N-dimensional hypercube, salted so as to be as far apart as possible, and then selecting instructions to maximize the mean "Hamiltonian-walk" distance to any useful one, still performing something nearly useless, but not entirely so (it should move just one step along the hypercube edge.) The winning entry would be able to implement and successfully run a bevy of realistic and diverse tasks (where the time required to execute can be meaningfully scaled) so as to maximize the minimum, hand (or highly) optimized code space required.

Some boundary conditions would need to be crafted, of course, for the contest.

Jon

I agree. It's the 4 term on the .tran line that most people miss. I mentioned this elsewhere. Spic will pick it's own time step based on convergence criteria then interpolate on your time grid if you only use the first two terms on the .tran line. This is normally fine since it makes the simulation run faster. But for accurate THD analysis, you need to insure measurements on actual time steps.

Your resolution is also a function of the time step. So your statement is accurate if you are thinking of the highest harmonic that you wish to measure. Obviously 4 points based on the fundamental frequency won't cut it.

I started experimenting with the circuit again tonight, paying particular attention to the feedback loop. I ran some simulations at high input frequencies and to my horror noticed this:

I ran a Bode plot to a much higher frequency than I had been doing, set things up so I could break the feedback loop and examine its output closely. It turns out my phase margin was virtually nil; it seems the bandwidth of the "untamed" cascode is just too high on its own. To rectify this situation I've put some capacitance from the output of the cascode into the base of the top transistor to give the amplifier a kind of dominant pole compensation. Checking the feedback loop again I tweaked it so I had a good 60 degrees of phase margin. The oscillation in the transient analysis now appears to be gone.

I see in another post you are from India. You should have learnt this always wise and correct maxims long ago.

"An engineer knows when a design is complete, not when there is nothing more to add, but when there is nothing more to take away"

and there is a close parallel story about an artist who is asked how he carves such a complex sculpture (for example an elephant)

"I just start with a block of stone and chip away everything that does not look like an elephant"

If your prepared to play I'll stay around and go step by step with you. Your purpose was to design a headphone amp.

Step 1 - write down what specs you want. (include voltage supply, power output, distortion and anything else you consider relevant)

I remember one published in ETI Australia pre 1980's - it used a quad TL071 all // - a worst choice for an op amp in this application is hardly possible, but in general it is a good approach and with an appropriate op amp an excellent one.

innews: snipped-for-privacy@q2g2000pre.googlegro

Yes. I also have tried odd numbers for maxstep such as 1.11u or

I remember seeing an instruction many years ago called HCF (Halt and Catch Fire).

I think it was actually a joke.

John

Sheesh, you can by a stereo Class D amp chip and add a minimal front end and direct coupled out, you can do this in a very tiny box. Sheesh.

What idiot chose a 15V rail? A headphone amp needs to be portable. If you are hooked up to something at home, you don't need a damned additional amplifier. It needs to run from a 6 or 9 volt rail, and if that requires stepping it up to drive the circuit, so be it. Another small circuit.

Plenty of headroom.

Class D is rarely HiFi and with headphones you can "shake the room" without shaking the room.

So get the 2 watt chip instead of the ten, and generally the MP3 based source 'signal' (let alone the engine that processed it) is worse than the circuit you made for boosting it.

So use a better audio chip. There are many.

Thanks for your reply. I'm not from India, perhaps you have me confused with another poster as I don't remember posting that anywhere - at least not while sober! They are good maxims, nonetheless.

Not simply a headphone amp, but a headphone amp that is unconventional in some way, that doesn't use a topology that has already been used in a thousand previous amps, whatever the merits and/or detriments of the common topologies used in those amps are. It would perhaps be fair to argue that by putting that condition on the project that I am not approaching it from the standpoint of an engineer, where, as your quotes succinctly put it, a design is considered successful when it as as simple as possible yet meets the design requirements.

I would counter that as a hobbyist one of the requirements is also that I have fun with the project, and to me building a project that is too similar to something that already exists isn't fun. I think it's also the process itself which is exciting and leads to discovery; how can anything new be discovered if not through being different for difference's sake?

As for electrical requirements, they are not stringent: a quarter watt into 64 ohms at maximum output seems like it should give plenty of reserve power into even large headphones, distortion should be as low as practicable, it should sound "good."

I would welcome any suggestions or guidance you wish to give.

BST - Backspace and Stretch Tape EPI - Execute Programmer Immediate

Like the Signetics 25120?

Are you going to wear something like this:

out for a walk?

Tube headphone amps must really piss you off, with their 300V rails.

Yeah. Except that I've had some of those memory problems.

If you want a topology not used before you are going to have to invent it BUT the most important part is that it *must* have some *advantage* over existing designs.

Something somewhat novel might be using a class A output with a constant current supplied by an LM317. Power is not an issue for a headphone amp.

Bitrex is the perfect example of what is wrong with "designing" with a simulator rather than designing "from the heart" :-(

Paste... paste... paste until it "works" will get you into trouble every time. ...Jim Thompson

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Interesting idea. Then all you need is a feedback triple... 3 to 4 trannies.

...Jim Thompson