Driving Your Speaker (& Not Damaging It) Just Got Easier

I had a discussion recently with a former business associate who was telling me about a recently released NXP speaker driver. He said it was a fairly sophisticated device intended to drive the speaker in a cellphone, tablet computer, or MP3 player. My initial reaction was, “Yeah, big deal, another 1W or 2W class-D amplifier. So what else is new?” He explained that this device is somewhat smarter than just a typical 50¢ 1W amplifier we were used to seeing.

He further explained that the problem that usually occurs with the small speakers is that 1) they are subjected to environmental temperature extremes; and 2) they are often driven too hard, which both overheats the voice-coil wire and stresses the mechanical attachment points of the speaker cone. So there is deterioration, both electrically and mechanically. So now I had a better understanding of the problem, but still no clear view of the solution. But that's where this new device enters the picture.

The new device from NXP is the TFA9887. It is a class-D amplifier, but it is also a lot more. For starters, it has a DC-DC converter so that its bus voltage is raised high enough to allow it to deliver 2.65W into a 4Ω speaker (external supply of 3.6V). It has a DSP on board that controls audio levels so that it can deliver the high power to a speaker that's rated for only 0.5W.

By monitoring the voice-coil resistance the IC can detect if the speaker is starting to get too hot. This avoids problems caused by excessive displacement of the speaker diaphragm. Not only does this improve speaker life, but it also improves the sound quality — for free air, closed box, or vented box applications. The device also detects load shorts and shorts to supply and ground.

The DSP also helps avoid clipping, which is an amplifier problem, not a speaker problem. But this is still a good thing — by carefully adjusting levels on the fly, sound quality is improved. It's clever enough to adjust levels as battery voltage falls and even prolong battery life towards end-of-charge. The DSP and its software are compatible with existing acoustic echo cancellers. There are no hidden license fees related to the DSP's software.

The device exhibits low susceptibility to RF interference — a nice feature in a device that operates right next to a small RF transmitter. This is due in part to the digital interface being somewhat impervious to input clock jitter.

The device can be controlled via its I2C interface, and audio is passed into the device via its I2S interface. The IC is housed in a tiny 29-bump package that's around 2mm x 3mm.

What I like about this device is that the guys that designed it thought about a problem that was well beyond the scope of just the IC. In this case, I had never thought about speaker characteristics and the way the speaker deteriorates with time. NXP did and provided a nice solution to the problem.

16 comments on “Driving Your Speaker (& Not Damaging It) Just Got Easier

  1. SunitaT
    February 25, 2013

    Brad, Thanks for the post.

    Buit-in micro speakers within mobile phone and other mobile music devices are limited to around 0.5 Watts, while the NXP speakers can deliver 2.65W using a technology similar to that found in hearing aids. This means moblie devices will enjoy the 6dB more volume when NXP speakers are incorporated into their build.

  2. David Maciel Silva
    February 25, 2013

    Probably not going to get all that power dissipate, because we have a high intake systems for mobile or portable equipment.

    When checking the data sheet not find a particular providing consumption to achieve maximum power.


    But its a good option with i2c protocol.



  3. eafpres
    February 25, 2013

    Note: edited from earlier (brain fade) post.  However I'm still wondering about the power consumption.

    I'm a little unclear on how the 2.65W drive would be used in mobile devices.  Here is an exmaple:

    iPhone 4s battery 1430 mAh

    2.65W speaker drive, generously assume 83% efficiency or 3W consumption

    Assume 5V, 3W = 5V*I; I = 0.6A

    30 minutes = (30/60)h*600mA = 300 mAh > 20% of battery life

    Or in other words, watching a short video would burn up a lot of battery just for the sound.

    Am I missing something here or is it unlikely that max volume would ever be used for any length of time?


  4. David Maciel Silva
    February 26, 2013

    I carried out some practice tests in the past with an mp3 player that I had, he had a autonimia more than 30 hours with a small cell, with a small resalva not turn the volume up more than 40%, thus the automia drops dramatically for only two or three hours maximum.

    Maybe an option would be a new audio amplifier IT


    Only 1.5mA per channel.

  5. Brad Albing
    February 26, 2013

    I'm certain that the key point here regarding battery life is precisely that the duty cycle of the loud sounds is pretty low. But if it is high, the intent of the IC under discussion is to prevent speaker damage and deterioration.

  6. mtripoli
    February 27, 2013


    Your calculation assumes a continuous current draw as if driving a purely resistive load. Audio is not that “simple”. Rarely would one draw that much current, and if one did it would be for very short (think “bomb explosion” for instance) period of time. For the most part, an amplifier sits at practically idle current when reproducing voice and soft background music. There are a number of papers regarding this that a quick search of Google should turn up. HiWave used to have a number of white papers talking about this; it looks like they aren't on their website anymore. TI is a good place to look. When dealing with Class D one must be careful about specs; they typically look much better than the “real world” in that they typically give values for highest current draw where the efficiency is best.

  7. Brad Albing
    February 27, 2013

    Yep – that's what I was trying to say, but you've made the point more clearly – and cited some good resource material as well. Thanks.

  8. eafpres
    February 27, 2013

    Thanks, mtripoli.  Audio is interesting; it seems to me that you end up with a bunch of (on average) over-designed stuff to deal with an occasional “bomb explosion”.  I guess if the users like it enough to pay for it, and that drives mobile sales and that drives content sales then this is all to the good for GDP expansion.

    In the early days of mobile/streaming video/audio people said it wouldn't work well due to the high bandwidth requirements.  Then the market entered a phase where low-res video on a (sort of) smart phone was fine; the attraction was the everywhere access not the AV quality.  As devices are getting better, it seems the pendulum swings back and demand for HD quality video and audiophile quality audio is coming back to smart phones/tablets.

    Considering how much I already spend every month on wireless services with data plans, I think I know who has the last laugh here!!

  9. mtripoli
    February 27, 2013

    People tend to want to lump everything together as if it were all apples and there are no star fruit. In the case of Class A amplifiers (or room heaters, take your pick what you want to call them), I don't know that you could “overdesign” the power supply. You hear some “outrageous” numbers that can be easily justified due to the topology of the circuit. In Class D, not so much. We've been conditioned to think that we all need 1000W+ amplifiers for “all” of our listening needs, when in fact most people would be surprised to find they don't “need” more than 10W or so. A large part of the “argument” for high wattage amps is the idea of “headroom” such that the amp never clips. Again, most would find that a great deal of clipping is tolerated by the average user and they never even know it. In the newer line of Class D (such as those available from TI) the devices have a “soft clip” circuit. People have exploited this and given it old terms such as “tube like sounding”. Makes one want to laugh out loud… 

  10. eafpres
    February 27, 2013

    @mtripoli–hmmm, surface mount tubes–can I get that in an 01005 package?

  11. Brad Albing
    February 27, 2013

    I'd like that #43 tube (or the newer version, the 6F6-G) in a surface mount, please.

  12. eafpres
    February 27, 2013

    Hi Brad–here you go in 4 steps:

    1) Go to the National Valve Museum

    2) Copy the picture of the 6F6G

    3) Paste into paint

    4) Scale to 1%

    Should fit into at least an 0402 and maybe 0201 layout.

  13. Brian Dotson
    February 28, 2013

    Hi, Brad,

    This is really cool! Not only does it monitor voice coil temperature, but the data sheet says this:

    “Adaptive excursion control – guarantees that the speaker membrane excursion never exceeds its rated limit”

    So, it's protecting the speaker in two different ways! I am really interested to know how they sense excursion limits. Does the back-EMF from the voice coil change when the voice coil hits the excursion limit?

    On a related note, in higher-power realms, I am really looking into another technology that maximizes the amount of bass energy that you can get out of small speakers from Audyssey Laboratories. It's called BassXT:

    Rather than sensing speaker conditions, this technology is an open-loop DSP process relying on careful characterization of the speaker system. As nearly as I can understand it, they:

    (1) Map the maximum power your product's speaker system (driver in the enclosure) can handle without hitting excursion limits or other distortion-causing events as a function of frequency. This is a complex function, as you might can imagine.

    (2) Real-time, the DSP runs an adaptive equalization algorithm that follows the music program information and maximizes the bass energy delivered to the amplifier stage.

    The result is serious bass extension. They have this little iPod dock product that is an excellent demonstration:

    We bought one of those at work to study what could be done. They have a pair of 4″ drivers in that little guy, and there is audible bass response down to 40 Hz! Notice I did not say that was a -3 dB response, but it's still quite amazing.

    Now, that being said, I would sure like to see higher powered integrated Class D amp solutions that had the real-time sensing you speak of. Those of us who sell multi-zone amplifiers to be installed with TBD speakers by the customer have to have the smart, closed-loop excursion control that you described in the TFA9887. Alas, market quantities drive the integration, and things that go in cell phones are far more important to manufacturers than amp ICs in the class I mentioned. I generally get a lot less integration in my world.

    It would be fun to figure out how to do this myself with external circuitry, though 🙂

  14. Brad Albing
    March 5, 2013

    To one of your points – yes, monitoring back EMF can be useful and predictive. Just like with a motor that is starting to stall, when the speaker voiuce coil starts to stall, back EMF drops.

    In applications where you're building a lot of the same things (e.g., cellphones), you have the additional advantage of being able to characterize the speaker (at least somewhat). That makes it easier still to predict when you are approaching its limits.

  15. Brad Albing
    March 5, 2013

    Oh – bass response. Almost forgot – one of the tricks you can do (and you likely know this) is you can filter out the fundemental tone from a complex waveform (like a musical note) but leave in the overtones (that the little speaker will have an easier time reproducing). Your brain will perceive the fundamental as being there even tho' it isn't. Auditory illusions.

  16. Brad Albing
    March 27, 2013

    Brian – with all the writing you did in your response, you could have written a whole 'nother blog. Which you should do. Looking forward to some more good ideas from you.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.