designing headphone amp

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

audiox

Well-known member
Joined
Feb 25, 2007
Messages
610
Practically all headphone amps in mixing consoles use push-pull emitter follower in the feedback loop to buffer the op-amp. I started to wonder why. With normal headphone sensitivity (1mW for 100 dB spl) you need about 11mA current for 100 dB sound pressure if headphone impedance is 8 ohms. Any op-amp can give that.

Is the low load impedance itself a problem if the op-amp can give enough current?

Of cource you can use a series resistor because the voltage required is also very low but I am interested in the case when it is not used.
 
Is the low load impedance itself a problem if the op-amp can give enough current?
Open-loop gain heavily degrades for standard opamps connected to such a load and crossover distortion increases (even if loop gain remained constant). As a guideline you might want to look at open-loop output impedance of the driving amplifier; it should be much lower than the load impedance--not easy at 8 Ohm...

Samuel
 
The more common mistake made by inexperienced console designers is to not properly deal with headphone amp layout. The relatively heavy ground path current between the headphone jack and PS can easily lead to crosstalk problems inside a console master section, if grounds are not well managed.

Regarding the transistor buffers, headphones come in a broad range of impedances so the buffer accommodates more models.

JR
 
It is certainly true that headphone amps often use an IC driving a low current complimentary pair of transistors, often because this is an inexpensive and reliable method of doing the job.

In some Soundcraft designs they simply parallel a couple of IC's (use both halves of an NE5532 for one channel for example) - to give more drive current.

It is also possible - and sensible - to consider using a power op amp IC, such as a Sanyo LA6500 - which will enable multiple headphone outputs to be used if required. This is a single channel IC - but a dual version is available.

Lots of ways to do this job!

JG
 
My 32 ohm cans sound a bit weak without at least 100ma current.

Buffer chips (buf634, LH0002) are real easy to put the loop of an opamp = 2 chips, and a few resistors.

check:
Head-Fi
Headwize
message boards for zillions of DIY schematics.


John said:
The more common mistake made by inexperienced console designers is to not properly deal with headphone amp layout. The relatively heavy ground path current between the headphone jack and PS can easily lead to crosstalk problems inside a console master section, if grounds are not well managed.

Would this be where a buffered ground channel would help ?

=FB=
 
> With normal headphone sensitivity (1mW for 100 dB spl)

That's high. I usually assume 92dB SPL/mW +/-10dB.

> you need about 11mA current for 100 dB sound pressure

But you "need" peaks to 105dB SPL for "realistic" sound level, and more for rock-n-roll, fine monitoring, or deaf musicians.

I happened to have 87dB SPL @ 1W speakers, a 50W stereo, and some piano recordings, in the same room with a piano. The live piano played louder than the stereo could without clipping.

You need 10mW-50mW power in the cans or you won't be happy. I design for over 250mW (but won't expose myself to that for more than a minute per lifetime; it is about "knowing" that any clipping is not in the HP amp).

50mW in 32 ohms means 55mA peak. Few "common" chips will do that, or not at low distortion, even if you build-out to the several hundred ohm minimum load for good gain performance.

Then you need a second jack for your asst. engineer or girlfriend, 100mA.

People do get-by with single opamp phone drivers. Works good for >32 ohms and modest demand. But if you are going to work around live music, and make critical decisions, on live music, you really-really want "ample" power, not "just adequate" power. The buffer, in quantity production, is about a buck of jellybeans and PCB space. On anything more than a $99 nano-Mixer, the cost is justifiable for fewer customer complaints. At DIY pricing it may be $5 of parts and $10 labor-equivalent, still cheap in context.

If you want Really Ample Power For Any Likely Headphone, ponder this essay. It assumes headphones are made to be driven to MAX rated power (else why build them so rugged?) and that you might sometimes want that max output. In many cases, this gives ~~120dB SPL, which is LOUD. But not out of line with control room speaker systems, which typically approach the output of an A-7 (117dB SPL max) with hopefully smoother sound. If your speaker tastes run more to Minimus 7 and 10W amps, then you could design your headphone amp to 20dB less, 1/10th the voltage/current. However the "cost of headroom", at HP power levels, is so much lower than the cost of loudspeaker gear, that we prefer to be a bit more generous.
 
... ponder this essay. It assumes headphones are made to be driven to MAX rated power (else why build them so rugged?) and that you might sometimes want that max output.

...in the odd event you like to monitor your headphones whilst they hang 'round your adam's apple ?

:green:


=FB=
 
We have a simple, good headphone DIY-project coming up sometime soon - it's under final testing by a couple of select local diy'ers:

all info, including layout as .pdf - http://www.gyraf.dk/gy_pd/hpamp/

image - http://www.gyraf.dk/gy_pd/hpamp/sch_small.GIF

Jakob E.
 
There is a new National chip that might be useful for headphone amplifiers.

National LME49600

See the data sheet at:

http://tinyurl.com/6jrlsb

ZAP
 
Having good voltage swing, low distortion, is also good for the talent. It's hard to sing when there's clipping in the cans. At one time I had a little 'Apex' 4-channel headphone amp. It used TL072's or something like that to drive 'phones. The distortion was horrible at anything other than a whisper, and I was running the 600 ohm K240's.

I did a DIY 4-channel, using a TI headphone amp chip, TPA6120, and I like it. It sounds good. Follow the layout they give exactly. It's think it migh be best to use SMT resistors on this one because of the speed of the amplifier.

-Dale
 
I would remove R6 connecting base of T2 directly to opamp's output. Max swing would drop down insignificantly, but pattern of distortions will be better. If layout is good C5 and C6 probably may be removed. C3 in my mind has too small value to shunt 220 uF electrolytic.

[quote author="gyraf"]We have a simple, good headphone DIY-project coming up sometime soon - it's under final testing by a couple of select local diy'ers:

all info, including layout as .pdf - http://www.gyraf.dk/gy_pd/hpamp/

http://www.gyraf.dk/gy_pd/hpamp/sch_small.GIF

Jakob E.[/quote]
 
[quote author="dale116dot7"]I did a DIY 4-channel, using a TI headphone amp chip, TPA6120, and I like it. It sounds good. Follow the layout they give exactly. It's think it migh be best to use SMT resistors on this one because of the speed of the amplifier.[/quote]

I did a stereo headphone amp based on the 6120 also and I like it. I agree about following the layout rules, plus make sure you have a large ground plane to wick away the heat. It may not be the easiest thing for the DIY crowd, as you need to spin a PCB so you can attach the thermal pad to a heat spreader/ground plane. When assembling the board, it helps to put solder paste on the thermal pad and use a heat gun. I used 0805 resistors and caps, too.

The distortion is pretty low and it sounds good with my Sony 7506 and 7509 cans.

-a
 
I will apologize in advance if this is in poor taste but there are several tiny tweaks that would on paper at least make a difference.

While I'm not convinced the LEDs in place of typical diodes to bias the output stage will be a significant source of distortion, as drawn that drive bias string will be starved for current well before clipping, and prevent the drivers from saturating to rail (perhaps not a bad thing for recovery time). Adding caps in parallel with the LEDs will provide more drive current and increase output swing.

------

Cap coupled gain stage and input presents unbalanced DC impedance to + and - opamp inputs. This will cause a DC offset voltage due to bias current times different resistances to ground. Making R1= 22k will provide first order DC balance (perhaps increasing value of r4 to 100k or splitting the difference. )

Adding a small resistor in series with the output (a couple ohms) will prevent output from dumping full current into a short circuit.

Likewise adding diodes backwards from bases to output on both drivers will provide current limiting but may be too low current for very low Z cans.

--------

33pF compensation cap is sacrificing a few V/usec of max slew rate. The decompensated opamp will not be happy with 100pf feedback cap. A smaller cap from opamp output to - input could be used instead with perhaps a small C to ground at - input (IIRC there is a couple pF in that package already). If an audio path LPF is desired a C could be added at opamp + input to ground.

Note: the schematic as shown with 100pF feedback and unity gain compensation will be better for rejecting RF coming back into the output line.

---------

Consider adding .1 or .01 uF disc caps across PS rails for low Z at HF.

-----

FWIW this is all nit picking... The design as shown should work and sound fine... but picking apart other's designs is what design engineers do for entertainment. Thanks for the entertainment.

JR
 
I don't remember what polarity of input diffamp transistors 5534 have; if they are npn C2 has to be reversed. But I'd use a metal film there.
Also, I agree with John: R3 should be 10K, R4 100K, or R2 decreased to 22K.
But LEDs are like rocks even on such a current, so caps would not decrease dynamic resistance visibly.
Max current is already limited by R5 and R6: if beta of output transistors is 100 max current when output shorted will be less than 300 mA, so no need to ask 8 guys to help to screw the bulb... But an opposite transistor will be killed by an opamp: too high reverse BE voltage... Couple of additional diodes connected backward will help (or couple of red LEDs, to indicate an overload as well).
 
> that drive bias string will be starved for current well before clipping, and prevent the drivers from saturating to rail

> Adding a small resistor in series with the output (a couple ohms) will prevent output from dumping full current into a short circuit.

> Likewise adding diodes backwards from bases to output on both drivers will provide current limiting but may be too low current for very low Z cans.


The large-signal output resistance is roughly (4700/Hfe)+47, or ~~100 ohms.

Picture the 5534 output flung full-up, output shorted. Output current is 17.3V/100R or 170mA. This is well within BD135/136 rating. If it held that DC level, the 47R would dissipate 1.7 Watts. Since the input is AC coupled we assume 0.9 Watts for a totally square wave and ~~0.5W for grossly distorted audio. If R7 R8 are 1/4W, you could lose a few, in odd situations, but it is unlikely and easily fixed with 1/2W parts. The DB parts are in no serious danger when the resistors are smoking.

With 32R load it approaches 17.3V/132R or 130mA, 4.2V peak, 250mW sine audio. This is roughly the max rated power of many 32R cans. Used wisely, sound is clean. (Used to abuse, you can melt cans... but that is true of any good practical audio system.)

For 300R load it can swing to around 280mW, a good value.

Jakob has approximated my "7V behind 29 ohms" guide for "very ample" headphone drive. I doubt he based his plan on my thoughts... more like he knows "what works and does not work" around studios, and found the similar answer his own way.

> you like to monitor your headphones whilst they hang 'round your adam's apple ?

250mW in headphones is no more excessive than 250 Watts in speakers. Most of us have used a Crown DC300, Phase Linear 400 or better for control room monitors.
 
Back
Top