I am doing personal project - portable MP3 player. Goals are sound quality + power efficiency. Price is not limited to a reasonable amount (100-200$ in parts is ok)

Will single-chip DAC+D-class amplifiers give me best possible power efficiency while having excellent audio quality? I've seen 24-bit ones but a little overpowered (like 1.5W per channel). This looks very juicy - No need to design analog part, all sensetive parts are inside the chip. Are there any pitfalls with this approach or with D-class amplifiers with headphones in general?

I afraid that classic op-amp based or AB-class amplifiers will eat too much power + I will need careful design of analog part after DAC.

For example: http://www.cirrus.com/en/products/cs43l22.html Separate output for headphones scares me.

Design targets:

  1. 32-50 Ohm headphones support @100mW
  2. LiIon power supply, the more play hours I get the better :-)
  3. Maximum quality (as long as I can hear the difference)
  4. I have 24-bit decoder output, so going down to 16 is possible to not that fun (extra analog filtering and noise shaping).
  • \$\begingroup\$ Is this to drive headphones? speakers? What is your power source? \$\endgroup\$
    – markrages
    Commented Mar 9, 2011 at 15:19
  • \$\begingroup\$ Just headphones (32 Ohm, up to some 100 mW), no speakers. Also I want to (optionally) be able to drive some 250 Ohm(also some 100-200 mW). Power source - LiIon+DCDC to get required voltages. \$\endgroup\$ Commented Mar 9, 2011 at 17:11

2 Answers 2


Your limiting factor is probably voltage swing.

If your driving higher end headphones your probably looking at a 250->300 ohm load, 300ohm for higher end sennheiser cans for instance. But some phones (mostly professional models) can go as high as 600 ohms.

At 300ohms:

For 100mW you need 5.47 Vrms = 15.48 V peak-to-peak ( 18mA RMS)

For 200mW you need 7.75 Vrms = 22 V peak-to-peak ( 25mA RMS )

You probably won't find an integrated part that is specifically designed for portable devices that can achieve those levels of voltage swing. Your best bet is probably an audio op-amp with a high drive current (TI makes several, as does analog devices) or a purpose built headphone driver, for instance the TPA6120A2 would fit your needs.

Your challenge is then to build an efficient power supply to generate the +-12-15V rails needed.

Not sure what your plan it for batteries but one way to make it easier and more efficient is to use two batteries, maybe two 9V batteries, center tap for ground giving you +-9V rails to start with. You end up with bigger/more batteries but you don't lose any power boost converting up to your + rail and then using a charge pump or similar to get your negative rail.

EDIT: I'd target 100mW MAX...200mW would explode your head with most phones. expect 97-102db sensitivity at 1mW for most higher end cans (could be lower for pro models). meaning at 127mW you'd be looking at SPL in roughly the 115dB to 120db range which is more than enough to cause hearing loss if listening for extended periods. Targeting ~63mW would put you at 112db -> 117dB which eases your voltage swing constraints and can still cause plenty of ear damage.

  • \$\begingroup\$ I see, ok, I officially ditch 250 Ohm support, only 32-50 Ohm range is left. I agree about 100mW, 200 were for some of the bigger & high-Ohm ones. \$\endgroup\$ Commented Mar 9, 2011 at 18:40
  • \$\begingroup\$ keep in mind when looking at headphones or speakers that the rated power is the thermal limit of the driver's voice coil, i.e. how much continuous power it takes to melt the voicecoil. Most of the time the driver reaches XMAX ( basically the drivers linear excursion range, i.e. its 'clean' excursion ) or XMECH (actual mechanical limits of cone excursion, 'bottoming out') well before it reaches its power handling limit. This is especially true when the driver is playing low frequencies(high excursion). the power needed to reach XMAX/XMECH is also dependent on the enclosure design. \$\endgroup\$
    – Mark
    Commented Mar 9, 2011 at 19:52
  • 1
    \$\begingroup\$ so unless your blasting a test tone in an SPL contest your limiter will usually be cone excursion. The driver will mechanically tear itself apart before the voice coil melts. This means that in most cases the power handling number should be a safety check in your design rather than treating it as a goal to reach for your amplifier design. \$\endgroup\$
    – Mark
    Commented Mar 9, 2011 at 20:09
  • \$\begingroup\$ Yeah, I am not going to listen 32 Ohm ones at 100mW, I bet comfort level is way way less :-) \$\endgroup\$ Commented Mar 9, 2011 at 21:28
  • \$\begingroup\$ @BarsMonster I'm a little confused, driving higher-impedance headphones at the same power would cause different SPL assuming same driver efficiency? I just got curious :s \$\endgroup\$
    – Sebi
    Commented Mar 22, 2016 at 21:43

I suggest you use one of the published high quality op-amp headphone driver circuits. The analog design of a class D amplifier is not straightforward. Class-D amps have approximately 0 dB of PSRR.

Have you figured out your power budget? I think you will find that the power savings of Class D is minimal over an AB opamp output stage. You might save more power by using a more appropriate DAC (one with no more bits than your source material.)

Also, a DC/DC converter will waste some power. With appropriate component choices, I think you should be able to power your circuit directly off a LiIon cell.

  • \$\begingroup\$ Well, while MP3 decoding would take about 30-70mW, I don't want to spend 300mW to get 100mW of output via opamp-based headphones driver. I do have 24-bit output of decoder, so while one can get decent 16-bit sound, this would require special handling of quantization errors. Power budget - the less the better, but not strictly limited :-) \$\endgroup\$ Commented Mar 9, 2011 at 18:38
  • \$\begingroup\$ Also, D-class chips I've seen does not require any analog circuit on output, so it can't be any simpler... \$\endgroup\$ Commented Mar 9, 2011 at 18:44
  • \$\begingroup\$ In my experience, class D requires more careful layout and design than an op-amp. Just like linear regulators are easier than switchers. \$\endgroup\$
    – markrages
    Commented Mar 9, 2011 at 19:13
  • \$\begingroup\$ At low power levels most modern class D amps are of the "filterless" variety. No external filtering is required and layout isn't critical for such devices. Class D does require more attention to the power rail(s) as gain is directly tied to the rail level, any dip reduces amp gain and thus impacts frequency response. You have to also be aware of bus pumping. For such low power these issues are pretty easy to handle though. Class D is more complicated to design for when your talking about 10's or 100's of watts. \$\endgroup\$
    – Mark
    Commented Mar 9, 2011 at 20:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.