8

I'm working on a USB-powered speaker design and could use some thoughts or design inputs from the community.

The data part is pretty simple - USB powers an I2S bridge that talks to an I2S amplifer. The power, however, is more challenging because I am trying to handle significant transients on the load while respecting the 500mA USB limit.

That pretty much translates to a current limit circuit that limits to 500 - (sum of all other chips usage) mA on the input. A large capacitance on the output of the current limit is meant to be the buffer to handle transient currents. This capacitance tied to a boost converter powering the amp should do the trick, right? Nope.

A boost converter's input voltage linearly affects the output current, and at the lowest input voltages, the output current is limited to 200mA on average for most parts I've looked at thus far. There is a compensation to be made in how low you set the voltage swing over the reservoir cap(s) vs output current of the boost.

The goal is to deliver the 500 - (sum of other chips) mA to the amplifier in normal operation, and adding up to 1 more Ampere with the aid of the reservoir cap and the boost circuit during transients. What I'm having a tough time figuring out is the right parts and the right capacitance for this design. I'm also willing to start the power part of the design over from scratch given a better solution.

Looking forward to hearing your thoughts on what a good solution would be.

Anindo Ghosh
  • 50,188
  • 8
  • 103
  • 200
Skateboarder
  • 136
  • 1
  • 5
  • Have you looked into use an inrush current limiter? It's a little bit of a different application - but it may work. I'm thinking your spike may be too fast for one. How fast are the transients? – Brad Apr 18 '13 at 20:33
  • Designing for twice the non-transient current level (*1 Ampere v/s 500 mA maximum from source*) to be sourced by reservoir capacitors will be challenging, to say the least. How much size can your design handle? Also, large capacitors will require further careful design compromises so they don't blow any protection existing on the USB supply line (or blow something more expensive) at initial charging transient. – Anindo Ghosh Apr 18 '13 at 20:34
  • @Brad: The design I have at the moment is using a current limited switch, which takes care of both inrush current and overcurrent. The transient time is an interesting topic. Since it's an audio application, you can say that the transient lasts in the hundreds of milliseconds because that's how long a loud crash cymbal or thumpin bass would drive it above the average. In the same sound clip, I can say the transients are at the tens of microseconds level because the peak to peak levels above RMS occur as such at the actual granularity of the signal, while the whole signal decays exponentially. – Skateboarder Apr 19 '13 at 00:00
  • @Anindo: Size is an issue, so I'm thinking of using the best dielectric possible for the biggest capacitance I can suck out of surface mount parts. All three dimensions have significant space restrictions - can't have gigantic electrolytic cans. =P The inrush current limit and overcurrent protection should keep the capacitance from screwing up the host circuitry. It's hard to figure out the magic capacitance for a design like this. Too much, and you're hurting yourself. Too little, and the job doesn't get done. – Skateboarder Apr 19 '13 at 00:04
  • So one thing I have been messing with (in another post) is creating a power well to power what I need to - and obviously putting a large (couple of) capacitors between it and an isolated ground. I then use an inductor to connect this power "well" to the main-power, to limit current spikes. So the inductor limits current spikes, and the cap gives it an on-demand "boost". The two together form a low-pass filter. I don't know exactly how to "tune" it, though. – Brad Apr 19 '13 at 14:21
  • As I understand your configuration places the storage capacitor on current limited 5V rail. Would a boost converter -> storage capacitor -> buck converter be maybe allowing for bigger voltage drop/swing on the storage capacitor (buffer function). – Rafal Powierski Aug 12 '13 at 14:43
  • Additionally could also use power more efficiently then possibly non-switching current limiter? If your audio amplifier is a D-class, the buck converter could be skipped, so how about placing the storage capacitor on the secondary side of your current boost converter? Boost converter input limit is often adjustable to match the coil saturation current ability. – Rafal Powierski Aug 12 '13 at 14:43
  • Basically it is all I = C dv/dt. I is the current being supplied by the capacitor. C is the capacitance. dv is the maximum allowable voltage droop. dt is the length of the transient. Put in the numbers you want and solve for C. If it is too big, you will have to try something different. I don't think an inductor will work well because you are dealing with 100's of ms. You would end up needing a large inductor with low series resistance. Rafal's boost-> storage cap-> buck idea would probably work better than trying to filter with an inductor. – user57037 Oct 01 '15 at 18:26
  • Shouldn't the large capacitance be on the *output* of the boost converter? – endolith Oct 02 '15 at 18:54

1 Answers1

3

Have you read Kollman and Betten, "Powering electronics from the USB port" ?

A large capacitance on the output of the current limit is meant to be the buffer to handle transient currents. This capacitance tied to a boost converter powering the amp should do the trick, right?

That sounds pretty reasonable:

USB power-->--[A]--|D>|-+--[B]-+-[E]--[S]
       |                |      |
       |               [C1]   [C2]
       |                |      |
      GND              GND    GND

where

  • [A] is a input-current-limited boost converter,
  • |D>| is the final output diode typical of such boost converters,
  • [B] is another DC-DC regulator (perhaps a buck regulator),
  • [E] is the audio amplifier,
  • [S] is the output speakers.
  • [C1] is an intermediate storage capacitor with wildly-varying voltage,
  • [C2] is a storage capacitor that B tightly regulates to whatever E requires.

What exactly is your question? I'm not sure, so I'm going to make 3 guesses:

Q: Is it possible to send 5 W(peak) to the speakers, while respecting the 500 mA USB limit?

A: Yes. While pesky laws of physics prevent us from sending a long-term average of 5 W to the loudspeakers while respecting the 500 mA USB limit (which limits us to pulling at most 500 mA * 5 V = 2.5 W at any instant), the idea of storing up energy during the "quiet" sections of music and then dumping that stored energy into the speakers during loud transients is reasonable. My understanding is that all mains-powered audio amplifiers that comply with EU-mandated power factor law EN61000-3-2 already do that -- they limit the current they pull from the mains wall outlet in order to maintain the appropriate power factor, and store the power in a bank of capacitors much like [C1]. During the zero-crossings many times per second, it's impossible to extract any power out of the wall outlet. So the designer sizes the capacitor bank big enough to supply all the energy required to keep the sound going until plenty of mains power is available again.

Sooner or later, it seems likely that some user will turn the volume knob all the way up and then try to play music that asks the amplifier to do something impossible -- to continuously dump 5 W of power into the speaker. I know some designers who would disagree on how the amplifier should respond to such an impossible request:

  • (a) Try to play the music perfectly at the human-set volume for as long as possible, until the storage capacitor is more-or-less completely drained. Then shut off the audio amplifier and stay silent until the storage capacitor is completely recharged.
  • (b) Try to play the music perfectly at the human-set volume for as long as possible; but set the clipping level proportional to the amount of energy available in the storage capacitor (preferably with soft clipping, also called gain compression); so quiet speech followed by very brief loud noises always play at the human-set volume, but loud sounds that go on for a while get progressively more distorted.
  • (c) Play the music perfectly at the human-set volume as long as the storage capacitor is close to fully charged; gradually adjust the actual volume quieter when it notices that loud noises have drained a significant amount of power from the storage capacitor; gradually return the actual volume level back to the human-set volume when it notices that the storage capacitor is fully charged. (Preferably make sure the capacitor bank is big enough and the volume is rampped down rapidly enough that the output is never clipped or silenced).

Q: Can I leave out regulator [B], and power the audio amp directly from the storage capacitor?

A: Unlikely. The voltage on that storage capacitor will likely go up and down a factor of 2 in normal operation. That's well outside the recommended operating range of typical audio amplifiers (unless they already have a DC-DC converter regulating their power), and no audio amplifier can perfectly reject such power-supply noise. Many DC-DC regulators can easily convert such widely varying voltages into something inside the recommended operating range of typical audio amplifiers.

Q: Is there a boost regulator that can put out more than 200 mA into a storage capacitor already charged to at least 10 V, while respecting the 500 mA USB limit?

A: No. I hear rumors that many USB devices are expected to work even when the USB voltage at the device is only 4.0 V. At 4.0 V, the 500 mA(max) limit gives a maximum power of 2 W. At 2 W(max) input power into a boost converter that has already charged a capacitor up to 10 V, the maximum physically possible current out of that boost converter is 200 mA.

Even though there may be a maximum of 2 W going into the storage capacitor bank, you can probably design that capacitor bank to easily supply 1000 mA at 12 V (12 W) during brief acoustic transients. (You'll probably want a bank of capacitors in parallel to lower the net ESR, or use low-ESR capacitors, or both -- rapid changes in voltage on a high-ESR capacitor causes it to heat up and fail).

Q: Is there some way to get more power to my USB audio amplifier?

A: Yes. Have you considered

  • using a separate mains power supply, similar to self-powered USB hubs?
  • using battery power?
  • using 2 USB plugs, one for data (and some power), and another plug just to pull more power from the host or a USB power port?
  • Somehow take advantage of higher current limit of 1.5 A in USB 2.0 (some sources say more; typo?)
  • Somehow take advantage of higher current limit of 3.0 A in USB Type-C (?)
davidcary
  • 17,426
  • 11
  • 66
  • 115