Minimal Headphones Amp

[john12ax7]

Try OPA1678/79 instead.

Why not OPA1641/42? More expensive but superior performance.

 

[thor.zmt]

Why not OPA1641/42? More expensive but superior performance.

Actually in listening (blind, preference) the tests results showed a preference for OPA1652/54.

This in turn is the same die as OPA1678/79 but with increased tolerances for DC and some tests omitted, so to speak the "B-grade" OPA1652/54.

And for mass producing low cost audio gear "more expensive" is the wrong direction.

Thor

 

[ccaudle]

This thread inspired me to look into the current boost stage for op-amp like @thor.zmt described in post 83.
One thing that was not obvious to me right away was what appears to be a bootstrap capacitor C19 from the output to the junction of the R43 and R52 resistors that provide the current path to negative supply for the Vbe spreader.
The output of the transistor stage should be following the output of the op-amp pretty closely, so what advantage does adding C19 have over just making C20 somewhat larger?

 

[ccaudle]

By the way, who told Thor he could go on vacation and stop checking his thread?

 

[JimJhn]

Not to hijack a thread or insert something unrelated, but...(and this IS about headphone amps I promise!)...

I just finished Mark Howard's excellent book "Listen Up", which chronicles the Canadian recording engineer, producer and musician's musical journey so far.

He has allot of really cool thoughts about recording music, and especially about how to get away from soul-killing sterile recording environments.

I really empathize with this, regardless of the differences in what areas of music interest me.

Having spent most of my daily life in commercial studios for more years than I care to enumerate (and I am not an engineer or technical at all), I've gotten to the point where the LAST place I want to make music in is a commercial studio. (EXCEPT for REALLY well-maintained places, and only if that's totally STELLAR, but I digress!).

One interesting thing he mentioned is that he buys old consumer stereo-systems to use as headphone amps instead of dedicated studio headphone-systems or the ubiquitous Crown DC-60s with distro or whatever, and it got me to thinking about that as I read this thread...

 

[kags]

One interesting thing he mentioned is that he buys old consumer stereo-systems to use as headphone amps instead of dedicated studio headphone-systems or the ubiquitous Crown DC-60s with distro or whatever, and it got me to thinking about that as I read this thread...

There are some producers, like Joe Baressi, who change monitors for each album they work on in the hopes that the new “lens” will help create a different outcome. You may or may not like this approach but it is certainly valid.

 

[ccaudle]

@thor.zmt I think you may have been on holiday when I asked previously; could you take a look at my question in post #143 and explain the purpose of the bootstrap cap from the output back to the Vbe multiplier stage?

 

[thor.zmt]

One thing that was not obvious to me right away was what appears to be a bootstrap capacitor C19 from the output to the junction of the R43 and R52 resistors that provide the current path to negative supply for the Vbe spreader.

The bootstrap capacitor acts as impedance multiplier. This means the load on the op-amp output is mostly the headphone impedance multiplied by output stage Beta.

With BD139/140 or BCP53/56 (same die) in the high beta group we have typically ~ 200 Beta. So a 16 Ohm headphone will appear as 32k Ohm load to the Op-Amp output.

This appears in parallel with the "tail" resistance to -V which is 2.4kOhm for ~5mA. If we split the resistor into 1.8k + 620R we load the output stage with an extra 620 Ohm load, but with the bootstrap and assuming ~ 1 Ohm open loop Z Out and a 16 Ohm headphone, the 2.4k become around 29k, so the load on the Op-Amp output becomes ~ 15kOhm, instead of around 2.2kOhm.

It also allows the negative swing to go all the way to the rail, while the positive side will always be limited to 1VBE plus other losses.

One could use a CCS instead of the bootstraped resistor, that has its own limitations and drawbacks.

Thor

 

[cyrano]

Thanks, Thor. Learned a lot from this one post!

 

[ccaudle]

The bootstrap capacitor acts as impedance multiplier.

Thanks for confirming. You do not usually seem hesitant about having an op-amp drive relatively low impedances, so I just wanted to confirm there was not something more subtle going on.

 

[thor.zmt]

Thanks for confirming. You do not usually seem hesitant about having an op-amp drive relatively low impedances, so I just wanted to confirm there was not something more subtle going on.

It depends on the Op-Amp, I usually use modern parts that do very into low impedances. In our case here it is simply to maximise the open loop linearity ahead of the power follower.

I found measurable THD reduction And increased maximum output with that cap in place vs absent.

Thor

 

[zzzzz]

sorry for bumping an old-ish thread, but since most of what is discussed throughout here are newer TI OPA series unity gain stable opamps, I wanted to chime in a question.

Rebuilding a headphone amp module from an Auditronics console (note: this one is noisy and has issues; actually in need of repair rather than just rolling opamps). Later versions used NE5534's but this one is old and has LM301's so there is room for improvement even just with swapping ICs and replacing the output transistor pair with simple BD139/140 off the shelf. I'm aware this is a very well known old scheme.

However, I'm thinking a path to a more "significant upgrade" would be to gut the board of everything in the signal path after the volume pot (or after C1/C4 and R7/R18? I guess) and just lay out a simple, minimal PCB with new parts and tack it on top with I/O wires run to their respective pads on the edge pinout.

A tiny little PCB module like that could then be retrofitted for various other applications.

The TI datasheets for OPA1655/OPA1656 have an application right there for a headphone driver using the new-ish BUF634A. Is there any room for improvement here? It looks pretty good to me... and just use one dual OPA1656? I don't think the output coupling cap (C3 and C6, 330uF) would be necessary anymore with the BUF634A? Maybe 47R resistor there instead?

TL;DR: I don't know a whole lot about driving headphones.





stock Auditronics headphone amp as-is with LM301. Just noticed the schematic notes 5534 as an option and doesn’t call for a compensation cap. Strange.

 

[ruffrecords]

The 5534 is internally compensated for closed loop gains greater than 3. The original circuit is set for a gain of 10 so the 5534 does not need external compensation.

Last time I looked one of the BUFs was very expensive but I think there was a much cheaper alternative. Just need to be sure to pick the right one.

Cheers

Ian

 

[ccaudle]

Last time I looked one of the BUFs was very expensive

I ran across that mentioned BUF634A recently and was shocked at how much less expensive it is than the original BUF634. Under $4 in singles, and the old design is about $15 in singles.

 

[zzzzz]

I ran across that mentioned BUF634A recently and was shocked at how much less expensive it is than the original BUF634. Under $4 in singles, and the old design is about $15 in singles.

Yup less than $4 in the USA from mouser or digikey which I feel like is great for what you get especially considering the price of the original

 

[Brian Roth]

With a slew rate spec of 3750 V/us, who can complain! <g>

Bri

 

[ccaudle]

With a slew rate spec of 3750 V/us, who can complain!

That is completely ridiculous for audio, but the nice thing about that is you can put it inside the feedback loop of any op-amp that you would reasonably use for audio, and just not worry about it. The bandwidth is so high you don't really have to think about changes to loop gain and how to compensate to maintain stability etc.
And it is nice to have just a few extra pins to solder (compared to e.g. discrete transistors plus biasing components, even though you do have to think a bit about thermal flow with the SOIC package, and the package with center pad is harder to solder).

On the one hand I hear "that is so easy to make a headphone amp, just throw down an op-amp, this buffer, and done!" and on the other side I hear Thor whispering in my ear "Bah! $4 for a buffer, I could make that with 62 cents of transistors and resistors, real engineering considers the cost/benefit ratio as well."
And on the other other hand I hear "you don't have cheap labor putting yours together, are you really willing to spend all weekend soldering down resistors just to save $3?"

 

[JohnRoberts]

That is completely ridiculous for audio, but the nice thing about that is you can put it inside the feedback loop of any op-amp that you would reasonably use for audio, and just not worry about it. The bandwidth is so high you don't really have to think about changes to loop gain and how to compensate to maintain stability etc.

Yes, that is literally how those ICs are intended to be used. The stability concern is delay/phase shift and indeed those are fast enough to be stable added inside most feedback paths.

And it is nice to have just a few extra pins to solder (compared to e.g. discrete transistors plus biasing components, even though you do have to think a bit about thermal flow with the SOIC package, and the package with center pad is harder to solder).

On the one hand I hear "that is so easy to make a headphone amp, just throw down an op-amp, this buffer, and done!" and on the other side I hear Thor whispering in my ear "Bah! $4 for a buffer, I could make that with 62 cents of transistors and resistors, real engineering considers the cost/benefit ratio as well."
And on the other other hand I hear "you don't have cheap labor putting yours together, are you really willing to spend all weekend soldering down resistors just to save $3?"

Headphone amps are as easy or difficult as you want to make them. Back in another lifetime I design a dedicated headphone amp (HB-1) that drove both legs of the headphone for 2x the voltage swing, 4x the power. This is arguably over kill. I design many for use inside mixers with more modest power.

JR

 

[ccaudle]

This thread has gotten fairly long, so for those coming in late this post from Thor provided an example of a discrete transistor buffer circuit:
Post with transistor buffer
And when requested Thor pretty generously provided some cookbook component values to use for 15V bipolar power supply:
component values for the previous circuit

An advantage of using the more expensive integrated buffer device is that integrated devices almost universally use current sources and current mirrors for biasing, so there is no supply voltage dependency. You can use an op-amp plus buffer circuit with a wide range of power supply voltages without having to change the circuit or modify component values.
With an integrated buffer you also get short circuit protection and thermal shutdown included, which would add a noticeable amount of complexity to a discrete transistor circuit.

 

[mhelin]

LME49600 is quite a common buffer used when building headphone amps (DIY). TI also recommends RC4580IP opamp as headphone amp (can source enough current).