Headphone Amp Redux

I will provide help while you design a new circuit. It will be your design. The current design must be simplified because it is too complex for the fact that it offers no advantages over existing designs - and it is also unstable. You will have to work to find most of those optimisations while keeping it as your circuit.

Your first step is to cut out the servo circuit. The reason is it does nothing practically useful. Replace it with a voltage divider as you describe. You might be able to optimise that to reduce the .3 volt offset.

If you can't find a way it doesn't matter - .3 volt offset is liveable and there will be more opportunities later

*Step 1.* Make the circuit work without the servo stuff

I don't mind if you say no - not interested or even if some point down the road you say it. Yes and No, dribs and drabs will piss me off. I'm disabled. I don't get paid for my time, but don't mistake that as "free"

al

e

he

he

to

way

ne

t - y , . G

I really don't think you guys get what Bitrex is trying to do. I assume he is working on a high end headphone amp. High end stuff tends to be all discrete designs. Parts count doesn't matter if you are building high end or just one of a kind.

Rather then eliminate the servo amp, I would make it op amp based. Technically the servo is not in the signal path, so it doesn't have to be discrete. Most likely the COTS op amp used in the servo would have less offset than any long tail pair you put together. You would pick the op amp based on offset rather than bandwidth since the servo is just an integrator forming a low frequency high pass.

I still suggest reading Randy Sloan's or Doug Self's amplifier book before rolling your own. [I've only read Sloan's book, but have read Self's papers.] A book is far better than a schematic plopped down on usenet that somebody proclaims to be good. [Hey, we know all good audio is designed in California.] Circuitry has philosophy. A good book or paper will explain the philosophy behind each stage in the amplifier.

There are DIY audio groups online. Also Nelson Pass has a website with tips.

Some podcasts with amplifier designers:

I wasn't particularly impressed with the "Krell" designer. He seemed like quite a hack.

I get the idea. The servo bit just isn't useful and needs to go. He might choose to stick one in latter, but he needs to get to a properly working amp first and *then* minimize the errors.

inal

s

be

The

m

the

o to

nyway

done

but

s -

ely

o s

vo,

64 t

th

x

it

set..

NG

t

e"

If I were designing the amp, at the spice stage, the servo would be made out of ideal components. Get the amp working well then put in a macromodel for the op amp. Of course, I'd be verifying in spice and not designing.

If you DC couple an amp, the servo isn't a bad idea. It can save your arse should some DC show up at the amplifier input. There is also the issue of turn on thumps, short circuit protection, etc. Mosfet outputs are much easier to protect without the circuitry getting in the way of the amplifier.

[snip]

As in...

The single-supply solution I suggested early-on.

[snip] ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

      Remember: Once you go over the hill, you pick up speed

MakeNoAttemptToAdjustYourSet expounded in news: snipped-for-privacy@4ax.com:

I believe the OP is trying some ideas from a "design" learning perspective. I see no problem with that. The learning aspect should be encouraged.

This is what is wrong with ham radio today. It's all too easy just to buy something already designed and made "better" than to do it yourself. Where is the fun and learning in that?

Warren VE3WWG

um, the output is feed into a 64 ohm set of headphones. It's not going to "blow"

Just because you select a more advanced, even lower part count solution, it does not mean that you cannot learn every aspect of how any given circuit/sub-cicuit performs/operates.

There are those that put things together to use them, and there are those that put things together to claim to be learning little nitty gritty particulars that they would LIKE to claim that others do not see or learn from their particular frames of reference.

I'm sorry, but that is flawed logic on your part.

I do not need to learn slow scan monochromatic transmission details to send my HD video feeds to my troops over Ethernet pipes. Someone already took care of all those details, and I lack nothing for not knowing every nitty gritty detail of a 10G Base link.

This is why man stacks his previous achievements upon his current set.

Otherwise, we would all still be trying to make a computer the size of a house using Fairchild ten transistor IC chips from 1960.

Instead, we are using CPUs that sport 2 BILLION transistor element, and that menagerie took years to develop and work toward.

QAM is a good example of how we start with a carrier and then learn how to pipe far more data into it than we even thought we could 50 years ago.

=A0 =A0 ...Jim Thompson

=A0 =A0| =A0 =A0mens =A0 =A0 |

=A0 | =A0 =A0 et =A0 =A0 =A0|

=A0|

=A0 =A0 =A0 |

Wrong. That op amp is in the signal path. I said the servo loop could use COTS since it is generally not considered in the signal path.

iginal

nd

is

to be

d. The

ohm

of the

rvo to

DC

anyway

e

by

e done

a but

ens -

imely

rvo

r

ngs

nd

ervo,

a 64

I

ent

with

lex

d it

es

ffset..

le

he NG

to

t

hat

nd

wn

'm

ree"

s o d h

Cables fail. You always need to assume an output will be shorted. This is the kind of engineering that separates a hobby project from something you can actually sell.

anyway

offset..

Seriously, stop making issues where there aren't any. There are a few ways to protect from a shorted output and price is the biggest dictator of what you can sell.

Beyond any of that. I am attempting to help Bitrex refine his design into something he can be proud of. At this point these "problems" you are raising have no relevance. At the moment he has an unusable and poor design - wraping that in another feedback loop that hides faults is ridiculous. You are not helping Bitrex.

It was as much commentary about the early days as joke. It included my favorite, "RMXG" randomize memory execute garbage; and many other "very useful" instructions.

KD5YI

and

believe)

Next to Whitney and across from Arapaho.

original

The

the

to

anyway

done

offset..

Don't let Miso get under your skin,,, I think he's gone Larkinesque, or worse, Slowmanesque :-) ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

      Remember: Once you go over the hill, you pick up speed

original

be

The

the

servo to

anyway

done

but

happens -

timely

servo,

64

with

it

offset..

NG

"free"

'Next Stop DIMBULBVILLE!'

--
You can't fix stupid. You can't even put a Band-Aid? on it, because it's
Teflon coated.

I'm sorry I haven't had a chance to respond to this thread over the weekend - I was working on my entry for the 555 timer contest that had its deadline today.

It has been suggested that this is a poor and unusable design, but since I made the tweak to limit the bandwidth of the input cascode somewhat it doesn't seem that way, at least from the simulation. I've tested it across a range of input frequencies and amplitudes and I can't get it to oscillate; the output appears to have very low distortion too. This is with both the AC and DC servo loop in place. The only way I can get it to oscillate now is at very low frequency if I reduce the value of the DC servo collector bypass capacitor too much. I'll give your suggestion of using an op amp for the servo instead a try and see how it performs then.

As I mentioned in a post below, after I limited the bandwidth of the cascode input stage I tested the circuit again at a variety of frequencies and amplitudes and I can't make it oscillate at all - this with both feedback loops in place. The output also has vanishingly low distortion if I've set the FFT up properly and it is to be believed. So, I'm not sure what's so terrible, aside from the "too many parts" criticism? I'm sure there's always the chance that it works in simulation but won't work in the real world with the properties of real components.

I'm sure I can design a headphone amplifier with fewer parts, but will the distortion be as low? Would it sound as good, since that's what's truly important? Of course, I have no idea if this circuit would sound good either as I haven't built it yet! Comparing the two circuits will make an interesting exercise.

I did - I even explained how to convert it to dual supply operation as an exercise as you asked. The post is somewhere up there.

Isn't attempting to create a circuit topology that hasn't been used before the antithesis of being "stuck in your ways"? How can one learn anything new if one only does what has been done before?

JT will pull your chain for his benefit, not for yours.

John