[Nemo de Monet]

Yeah, I know the idea of powering a headphone amp off of 5V USB power gets asked all the time, and is inevitably shot down by everyone as useless and impractical. Thing is, I don't necessarily think it is useless. Oh, sure, you can't power even the tiniest tube amp off of USB power, and 5V isn't a really useful voltage even for solid-state amps. That said, a DC-DC boost converter will give you, starting with a USB port's 5V, an usefully high voltage at a few dozen milliamps. Maybe not enough to power a tricked-out Pimeta, but enough to run a fair number of less power-hungry amps, like the Mini^3, just as one example.

I did some looking around, and settled on the Texas Instruments TPS61040/TPS61041 devices, primarily because they come in a five-leg SOT-23 package that can be soldered fairly easily by hobbyists. The datasheet includes a couple of reference designs - including one that produces 18VDC output from an input between 2V and 6V DC. See where I'm going with this?

At 80% efficiency - which the reference design achieves - 5VDC in at 100ma will get you 18VDC out at about 22ma - enough to power, and optionally charge, an extended-runtime Mini^3, or power most CMoy designs. If you were to run the design from a powered hub, capable of providing a true 500ma at 5VDC, you could get a maximum of right around 100ma output at 18VDC - more than enough for most portable, solid-state headphone amps.

Yes, the whole doohickey is a switching power supply, and undoubtedly produces an output voltage of less-than-audiophile-perfection. Given that there are things you can do to minimize the ripple and noise - and given the quite high PSRR of most op-amps - I don't think this really an issue for any but the obstinate purists. Further, I suspect that most people are willing to compromise theoretical, on-paper quality for real-world convenience.

By using 0805 and larger SMDs, and a B-type USB jack, the whole thing can be made to fit in a 24 x 35mm space (as wide and tall as, but a fair bit shorter than, a single 9V battery):

(Ignore the silkscreen problem, doh.) I deliberately used a large, though common, inductor on this prototype to allow flexibility - with a little creative effort, a number of physically smaller inductors should be able to fit the pads. There are also wirepads to easily accept input from a non-USB source - if you wanted to recycle an old 5.5V cellphone charger, for example, you could easily wire up a power jack, or whatever. Total parts cost, excluding PCB, is probably under $10 USD ($2 IC, $2 inductor, $0.50 diode, five resistors or capacitors at a total of $2.50 or less, USB connector at $2 or so).

I should have a prototype board in a couple weeks, and then I'll see whether USB power really is as viable, within limitations, for powering small headphone amps as I suspect it is.

Thoughts, observations, criticisms?

 

[DKJones96]

If you are going to run off of a USB 2.0 powered hub it is possible to get 1 amp. 500ma is the specification for a standard USB device, but if the it does not actually communicate with the USB host it is recommended that it only draw 100ma or less to avoid causing problems.

I made my cmoy usb powered but between the noise coming from the sound card and the low-profile power supply it was horrible to listen to. My cmoy will work with as little as 2.7 volts (before rail splitting) full volume without clipping.

 

[MisterX]

The most obvious question is:
How are you planning on mounting this "doohickey"?

Other comments are:
Dual "C1" labels.
Mixed up routing, you have a big fat trace running from C1 but a skinny trace running from Vbus, shouldn't it be the other way around?
You could clean that all up by running the Vbus trace directly to C1.

And last but not least your ground plane is not continuous, the ground sides of C2 & R2 are only connected via the thermal pad on IC1.

 

[Nemo de Monet]

Quote:

Originally Posted by MisterX /img/forum/go_quote.gif The most obvious question is: How are you planning on mounting this "doohickey"?

At the moment, I'm not.

It's a proof-of-concept, not a finished product. I had envisioned a couple possibilities:

1, As a USB-powered device, as a slightly larger overall board fitting into an itty-bitty case, a la the BantamDAC, as a sort of "USB wallwart", for powering whatever;

2, Also as a USB-powered device, internally mounting in the space in a much larger case normally occupied by a 9V battery, probably held in place by fairly aggressive foam tape (There are no components on the bottom of the board, and this is a proof of concept, remember - it wouldn't be too hard to swap the through-hole USB jack for a surface-mount one). If there was any interest in using (something like) this in lieu of a battery in a Mini^3, it wouldn't be too difficult to add a hole to line up with the "ribbon" hole on the Mini^3 board, and the two could then be mated with a small nylon screw and nut;

3, As a plain old power converter - wires to a jack in, wires to whatever out - again, mounting would probably be with adhesives, since the whole thing wouldn't weight more than, what ten or twelve grams?

Quote:

Other comments are: Dual "C1" labels.

Yeah, I know, I know.

Quote:

Mixed up routing, you have a big fat trace running from C1 but a skinny trace running from Vbus, shouldn't it be the other way around?

Yeah, probably. The auto-routing was all screwy, and I tried to fix a bunch of stuff by hand, but obviously missed a few things. Repeat after me: proof of concept, proof of concept...

Quote:

And last but not least your ground plane is not continuous, the ground sides of C2 & R2 are only connected via the thermal pad on IC1.

That was actually sort of intentional. I had originally been looking at using another chip - an MSSOP-10 Intersil part - whose design notes recommended fairly strongly that the power and signal grounds be separate, and only connected at one point. The TI datasheet, by contrast, suggests either a star ground or a groundplane, and I'd been using, for no real particular reason, that area as the star ground before I decided "what the hell" and added a ground plane.

What can I say, my layout skills are teh suck. Who knows, though? The doohickey might even work!

 

[rds]

Quote:

Given that there are things you can do to minimize the ripple and noise - and given the quite high PSRR of most op-amps - I don't think this really an issue for any but the obstinate purists.

Ripple and noise are definitely not the only problems. The idea that this power supply is good enough for all but obstinate purists is pretty absurd.
...anyways...
You should isolate the output at least.
Also throwing an inductor on the output with no resistance or capacitance is kind of useless.
It would be convenient if you'd post a schematic so your design is more clear.

 

[amb]

A big problem with using USB power is the ground. If you're going to use the computer also as the analog source (either via its onboard sound or an external DAC), or if you use any other source that has a 3-prong power cord which has its AC ground tied to the same ground as the computer, and then you connect it to a virtual ground amp (Mini³, M³, Pimeta, PPA, cmoy, etc), then the amp's virtual ground will become shorted to USB power's ground, which is the amp's negative rail. Bzzzt -- dead TLE2426.

 

[Nemo de Monet]

Wierd... Head-Fi ate my last response to this thread. Damnit...

Anyway, yeah, I totally forgot the ground problem, doh. Would solving it be as simple as using an isolated DC-DC converter, or am I missing something else, here?

 

[amb]

Yes, to avoid the problem with virtual grounds you'll need to isolate the ground somehow. The current limitation of USB power is another issue to contend with.

 

[joneeboi]

There is a small handful of DC/DC converters at Mouser and Digikey that have well over 500V of isolation on a relatively small footprint. USB can provide 2.25W of power, so depending on the efficiency of your circuit, it can produce a good amount of power. Avoiding the use of USB extenders/hubs would also be good practice.

 

[Nemo de Monet]

IMO, the current limit is acceptable, and can be dealt with; yes, it's a limitation, but not really a major problem.

How about something incredibly simple, based on TI's DCP01/DCP02 family of isolated regulators, like the DCP010512? The 2W versions are pin-compatible, and probably of more interest for what we're looking at. A pretty basic implementation is like this:

And a 1" x 1.5" implementation, using 1206-size SMD parts, could look like this:

Depending on the chip used, 5VDC in from USB or another source would get you an isolated output anywhere from 9VDC to +/-15VDC, at efficiencies around 85%; using the DCP010512B, you'd get 12VDC out at around 35mA for 5VDC 100mA in, which any USB port should be able to provide, and it'd be internally limited to around 70mA out for 200mA in. A resettable polyfuse (1206 package again) would prevent overcurrent from the USB line. While the chips are limited to 1W/2W depending on model, the datasheets indicate they can be paralleled, basically like the BUF634, so people who know what they're doing could, if needed, build one to take advantage of the voltage options (+12, +15VDC) that are only available as 1W devices, while still taking advantage of the current potentially available from the USB bus.

There are SMD versions of the chips, of course, but they're no cheaper, and would be harder to parallel.

Might that be more workable?

 

[rds]

What is your design goal? Is this for a specific project or intended to a be generic commercial product (you keep mentioning what 'people' can do with it)?
Imo, this is not something that you design generically. You need to know exactly what you're powering and implement the design accordingly.
This is a good chip and I have a design to power a pimeta with 12V regulated and using lmh6321 buffers. I haven't yet had a chance to test it.

 

[Nemo de Monet]

Quote:

Originally Posted by rds /img/forum/go_quote.gif What is your design goal?

Principally, to demonstrate that USB is a viable power source for low-power applications, particularly portable headphone amps. Assuming that's the case, and there's any kind of interest in it, to provide as flexible and user-friendly a design for the individual DIYer to use or abuse as he or she sees fit.

Quote:

Is this for a specific project or intended to a be generic commercial product (you keep mentioning what 'people' can do with it)?

Probably a completely generic device, yes, hence the emphasis on flexibility. I'd suspect that, assuming the little doohickey works as intended, some interest might develop in specific applications for the design (like a USB charger for the Mini^3), but that's, IMO, immaterial.

Quote:

Imo, this is not something that you design generically. You need to know exactly what you're powering and implement the design accordingly.

I disagree. IMO, something like this - assuming it works as intended, obviously - could be used to provide power from a computer for any device that needs a single or split source of <=30VDC at 2VA or less, which covers a few things, including most portable solid-state headphone amps. It's flexible enough that it provides the potential builder - who I tend to generically refer to in short form as "people", sorry of that somehow bothers or offends you - a quite large number of options. (From USB or another 5VDC source, depending on the chosen chip, output could be +5VDC, +/-5VDC, +/-6.5VDC, +7VDC, +9VDC, +12VDC, +/-12VDC, +15VDC, or +/-15VDC.) It could fit into a little plastic case like that used on the BantamCableDAC, and become a semi-custom USB dongle to power and/or charge a Mini^3, or whatever; it's basically the size of a 9V battery, so it could fit into a lot of device cases meant or used to hold the same. (USB-powered +/-6.5VDC CMoy in a mint tin, anyone?) Hell, with some of the available chips in the family, people - excuse me, "potential builders" - could ignore the USB part, feed it power from another source, and use it as a fairly tiny virtual-ground/rail-splitter circuit; a TLE2426 with higher current capability.

Then again, evidently I'm the only one who thinks this is potentially interesting, so what do I know?