Limiting Power of TPA6120A2 for Headphones

[PRR]

> Adding diodes somewhere

Ummm.... use a smaller resistor than 39 Ohms? (If you think you need 39r to the phones, use 2.5r for current-sense and 36.5r added outside the sensing.)

> doesn't have the thermal pad

Seeing as we have 0.25-0.4W max, and will normally run nearer 0.03W or less, not sure heatsinking is really needed. (That's class B. My class A headphone amps had BIG alloy extrusions yet ran uncomfortably hot.)

 

[ricardo]

Then again a powerful op amp like NJM4556 falls short (and it doesn't have the thermal pad on the bottom).

Wanna tell us on which phones 2 x NJM4556 on 2 x 9V like http://nwavguy.blogspot.com.au/2011/07/o2-headphone-amp.html fall short?
BTW, it also has some very sensible protection.

 

[abbey road d enfer]

I just looked at the datasheet of this headphones amp. There is not one single mention about interaction between the output resistors and the drivers. With most of the modern drivers, the value of 39 ohms may induce frequency response error of up to several dB at LF. Is it really consistent with claims of "High Fidelity"? Unless specified otherwise, most h/p's are tested with a near-zero source impedance; the published frequency response reflects these test conditions.
Check http://nwavguy.blogspot.fr/2011/02/headphone-amp-impedance.html
And is a slew-rate of 900v/us a necessity (about 2ns rise-time in typical conditions), considering the precautions needed to make the circuit stable?
I know it's petty bickering, because the musicians at the receiving end have already shot their ears with heavy doses of overdriven Marshalls. 

 

[Bo Deadly]

Wanna tell us on which phones 2 x NJM4556 on 2 x 9V like http://nwavguy.blogspot.com.au/2011/07/o2-headphone-amp.html fall short?
BTW, it also has some very sensible protection.

If you parallel two NJM4556 together, yes, that would work pretty well and that person seems to have thoroughly proven that. And since the TPA6120A2 requires a buffer to keep offsets low it would probably take up about the same amount of space.

Although I'm not sure if I'm reading the schematic correctly. Are they using 1 ohm on the outputs? With the offset of a bipolar amp that could burn a few mA for no reason (or make it sound better if you like class-A). Those should probably be more like 10 ohms.

But what is this "sensible protection" you speak of?

 

[Bo Deadly]

I just looked at the datasheet of this headphones amp. There is not one single mention about interaction between the output resistors and the drivers. With most of the modern drivers, the value of 39 ohms may induce frequency response error of up to several dB at LF. Is it really consistent with claims of "High Fidelity"? Unless specified otherwise, most h/p's are tested with a near-zero source impedance; the published frequency response reflects these test conditions.
Check http://nwavguy.blogspot.fr/2011/02/headphone-amp-impedance.html

Maybe this is what benb was talking about?:

I'd consider a resistor divider on the output of the amp, with the bottom resistor (or at least their parallel equivalent) scaled to be no more than 10 percent of the headphone impedance. This gives good damping, which some headphones (thinking of my Sony MDR-7506) really need.

So maybe a ~10 ohm bottom resistor in parallel with the phones this would significantly reduce frequency response errors yes?

And it would also reduce load variation.

UPDATE:

Damping resistor depends too much on phones impedance. With high Z phones it will shunt too much and with a higher damping resistor it won't have any effect on lower Z phones. But for specific phones it would work and has a nice side effect of limiting power.

 

[Andy Peters]

If you parallel two NJM4556 together, yes, that would work pretty well and that person seems to have thoroughly proven that. And since the TPA6120A2 requires a buffer to keep offsets low it would probably take up about the same amount of space.

You could use a pair of OPA551 in DIP package and be done with this.

 

[abbey road d enfer]

You could use a pair of OPA551 in DIP package and be done with this.

You don't believe in the superiority of current-feedback designs? With only 15v/us, the 551 is bound to sound awfully bad in comparison with the 900V/us of TPA6120 

 

[Andy Peters]

You don't believe in the superiority of current-feedback designs? With only 15v/us, the 551 is bound to sound awfully bad in comparison with the 900V/us of TPA6120 

I've done the comparison. The 551 sucks eggs!!!
-a

 

[clintrubber]

I want to try TPA6120A2.

The problem is this chip is a little too powerful for headphones. It's nice to have that power but it scares me a little to think I am attaching speakers to my ears with an amp that can put out more than 1 watt per channel. Eventually something is going to break and I don't want to loose an ear in the process.

Hmm, I have an Aphex headphone amp upcoming that uses this chip as well (in fact, one for each of the four outputs), and I can understand your concern. I've seen the potential sloppy soldering being mentioned for this model (454), so I guess I'd best check it. Thanks for bringing up this 'worry'. Will have a read through this thread - what have you ended up with meanwhile ?

Regards

 

[Tubetec]

I know canford make passive headphone limiters for hearing protection in Tv and radio studio applications ,goes between the amp and the headphones ,the limiting level is adjustable ,you might be able to find the schematics on the net.
There really is nothing worse for a musician or vocalist than a big ugly distorted crack through the cans ,transistors and op amps tend to distort very badly when over driven ,large amounts of very high order harmonics are created . Ive seen it happen in sessions where hours if not the rest of the day was wasted while the reciever of the ultrasonic shockwaves ears recovered .For my own setup I built a dedicated El84 headphone amp ,I used some small transformers salvaged from a stereo dansette type record player ,wired up the El84's as triodes and added a proper ht supply ,delivers around 1.5 watts max per channel into 4 ohms  ,and even if it gets over driven the resulting transient is way less harsh than transistor clipping . Of course pure class a tube single ended is very power hungry but for me its a price worth paying just so the listener never has to deal with supersonic ear splitting harmonics.

 

[Bo Deadly]

Hmm, I have an Aphex headphone amp upcoming that uses this chip as well (in fact, one for each of the four outputs), and I can understand your concern. I've seen the potential sloppy soldering being mentioned for this model (454), so I guess I'd best check it. Thanks for bringing up this 'worry'. Will have a read through this thread - what have you ended up with meanwhile ?

Regards

I actually ended up doing a standard parallel NJM4556 type circuit (added to my panel of boards at the last minute because of this discussion). I have the TPA6120 boards as well but I never even tried it.

Honestly I have sort of been thinking that a regular op amp with follower transistors would work just fine:

At first glance one might think that this circut is lame because the bias resistors are going to significantly limit the output swing. But is this not a good way to limit power? I haven't done the real math but even if output was limited to only 5V, with 600 ohm phones thats 42mW which would be loud enough for most folks. As for distortion, how much distortion do voice coils introduce? Probably a whole lot more than even the worst op amp + transistor buffer.

 

[clintrubber]

For my own setup I built a dedicated El84 headphone amp ,I used some small transformers salvaged from a stereo dansette type record player ,wired up the El84's as triodes and added a proper ht supply ,delivers around 1.5 watts max per channel into 4 ohms  ,and even if it gets over driven the resulting transient is way less harsh than transistor clipping . Of course pure class a tube single ended is very power hungry but for me its a price worth paying just so the listener never has to deal with supersonic ear splitting harmonics.

It's a pesky topic, at least when going for max fidelity & detail ... as will have been stated previously we want to have blameless headphone amps to iron out any eventual flaw that might limit those headphones showing their potential - but at the same time we want to put a brake on that in view of safe listening levels & gear-error-consequences.

Using a tube-driver might indeed be a good one, if the output impedance can be kept low enough, which will be doable. The source impedance requirement is a debatable spec*, some say for instance the minimum series-R of 10 Ohm that is required by the TPA6120  is already too much - depends of course on the headphone load as well. (*: '8-1'  rule)

Regards

 

[clintrubber]

I actually ended up doing a standard parallel NJM4556 type circuit (added to my panel of boards at the last minute because of this discussion). I have the TPA6120 boards as well but I never even tried it.

Honestly I have sort of been thinking that a regular op amp with follower transistors would work just fine:

Thanks for posting. Such a hybrid probably more fun (& sure good enough) than applying a 'potentially temperamental ready made chip'.  Saw the related schematics of say the Rane headphone amps, but they use a different bias scheme.

I realize it remains a difficult aspect of the 'worry', if that worry is stated in terms of that any malfunction should be kept at bay to prevent hearing damage... I guess this posted circuit doesn't necessarily do that either... as in: guaranteed before the ear-bleed line.
Or if we could do a circuit that gives that guarantee, it might need re-calibration when swapping for a different headphone.

As for distortion, how much distortion do voice coils introduce? Probably a whole lot more than even the worst op amp + transistor buffer.

Another difficult one, the voice coil distortion might be more indeed, but probably less nasty of nature. This also relates to the tube-approach above.

Regards

 

[Andy Peters]

Honestly I have sort of been thinking that a regular op amp with follower transistors would work just fine:

At first glance one might think that this circut is lame because the bias resistors are going to significantly limit the output swing. But is this not a good way to limit power? I haven't done the real math but even if output was limited to only 5V, with 600 ohm phones thats 42mW which would be loud enough for most folks.

You could use an OPA551, which gets loud enough to clean the wax outta your ears.

 

[abbey road d enfer]

It's a pesky topic, at least when going for max fidelity & detail ... as will have been stated previously we want to have blameless headphone amps to iron out any eventual flaw that might limit those headphones showing their potential - but at the same time we want to put a brake on that in view of safe listening levels & gear-error-consequences.

It is indeed a rather complex subject, that cannot IMO be solved by starving the drive capability of the circuitry.
I think you should use a smart limiter.
Engineers at Shure, dbx, Sennheiser, Rane,..., have killed a few brain cells over these issues; if they had found a simple solution, they would have used it.
You may not need the complexity of a dbx IEM, so a 4301-based limiter using adequately shaped side-chain EQ should provide an acceptable solution. It is quite easy to assess, too; use a limiter with an EQ in the side-chain and you're good to go. When you reach a satisfactory setting, you just have to turn it into an integrated product.

 

[benb]

Oh, no, not THIS thread again ...

(not objecting to the thread, but rather the subject, which is indeed complicated and frustrating)

I'm thinking The Real Solution is a (chicken head knob)  rotary switch at every headphone output jack, with each switch position being a make-and-model of headphone. It's just about The Only Way...

Okay, add generic settings to match common (uncommon?) headphone impedances/sensitivities.

Thinking another moment, current could be measured with a small series resistor, telling (with the voltage) a microcontroller the impedance of the headphones, thus switching (presuming all headphones have similar sensitivities) the max voltage level to match. Or have the controller detect headphones being plugged in, do a frequency/impedance test (at moderate volume in the phones) and from that looks up the exact model phones that were just plugged in. (Yes, it's a "simple matter of programming" - an English description is so much easier to write than code)

It is indeed a rather complex subject, that cannot IMO be solved by starving the drive capability of the circuitry.
I think you should use a smart limiter.
Engineers at Shure, dbx, Sennheiser, Rane,..., have killed a few brain cells over these issues; if they had found a simple solution, they would have used it.
You may not need the complexity of a dbx IEM, so a 4301-based limiter using adequately shaped side-chain EQ should provide an acceptable solution. It is quite easy to assess, too; use a limiter with an EQ in the side-chain and you're good to go. When you reach a satisfactory setting, you just have to turn it into an integrated product.

That's much better than flat-out clipping at whatever the "Max dBSPL" setting is! This needs a light that comes on when limiting is in effect, so the listener knows they're not hearing the full dynamic range.  And add another chicken head knob for max dBSPL, maybe 85 to 115.