[joneeboi]

Greetings DIY subforum,

I am interested in designing a USB-powered headphone amplifier that fits with tooleAudio's BantamDAC in a Hammond Mfg. 1455C80x case. I have had this idea in my head for maybe a year, but flying solo, I haven't been able to get very far. My interest was tickled by rds' USB Powered Pimeta thread which brought up the topic of powering an amplifier with USB power. I got the brilliant idea to design a USB-powered headphone amplifier when my Toshiba laptop's physical volume knob was snapped off and music was impossible to get from the onboard sound card. I figured I'd make an Alien DAC and throw it into a 1455C80x with a USB-powered amplifier before interest waned when I got a new laptop and I got music again. However, thinking about it again has made me really want to do this anyway. I figured I should contribute my time and energy to this board for all I've gained from it, and I want to do it for fun as well.

First things first, how viable is a USB-powered headphone amplifier? Basically, I want to recreate my own version of the dsavitsk's HPDAC but with just the amplifier section. Maybe I'd be better off including my own DAC section, but there's a perfectly good module I can point to without spending time and energy unnecessarily.

Next, I outline my design goals in order of highest priority to lowest.

• USB-powered
• Fit into the Hammond 1455C80x
• 3-channel CMoy amplifier
• Low-cost
• Relatively easy to build

Items 3. through 5. aren't hard and fast rules, but I'd like to start there. I'm glad the BantamDAC came out when it did because it has a smaller footprint and costs less than the AlienDAC. I realize that I could also get something like the HeadAmp Pico, but I'm not interested in spending that kind money on my audio wares just yet. And despite being part of an audio forum, I left out "6. Great sound quality." I'm more interested in getting it to work first, and then I'll play with the idea of beefing up the amplifier. Unfortunate but that's what I have to work with at the moment.

I realize I'm walking alongside giants in this forum, so I'm open to learning from all of you. To be honest, it's a bit intimidating even starting this thread which is partly why I didn't start one for so long. I'm completely open to doing the legwork necessary for getting this idea out the door, I just need ideas first.

Now, let's talk specifics.

The biggest hurdle here, I feel, is power supply. rds wanted to power his PIMETA, but there is the tricky issue of the virtual ground versus "true earth" ground. This device will be lugged around with my laptop, so the earth ground will be whatever the laptop uses.

I was thinking of going with a 3-channel CMoy design. Most DIY SS amps I've seen run this course while using virtual ground, eg. PIMETA, PPAv2, M^3, Mini^3, so it won't be as easy as lifting the amplifier schematic from one of those designs and plugging it into a DC-DC boosting circuit. Buffers probably won't be implemented because of the already heavily loaded power supply. If there is room left over, maybe I'll consider it. Keep in mind that the BantamDAC will also be connected to the USB port, so the PCM2702 and TPS regulators will be drawing some current as well. I hope to have a circuit that circumvents the whole trouble of virtual ground vs. earth ground.

I'm not tied to any one set of opamps at the moment. I don't have much experience outside of OPA627, OPA2134, OPA2132, OPA2227, OPA2107, AD8320, AD8310, JRC4556, AD8397 and OPA690. Think CMoy, PIMETA, A47 and Mini^3. Any suggestions? Opamp rolling sounds tempting, but I'm worried the DIP package will take too much board space. Also, sticking to a certain set of opamps will help with containing current draw, but maybe that can be decided by the builder.

So that's about it for now. I really should be studying for finals at the moment, but that can wait just a bit. I'll be done all my semester on the 17th, but beyond then, I'll have plenty of time to work on it even into the new year as I am out on an eight month Engineering co-op work term. Any and all comments are welcome. I apologize in advance to those who are tired of seeing yet another one of joneeboi's threads with super huge first posts.

I foresee oodles and oodles more of me updating the original post.

 

[joneeboi]

[size=x-small]Presenting the
[/size][size=x-small] [size=small]Bus-Powered Mini^3

[/size][/size]

[size=x-small]
In theory, this circuit is supposed to be identical to the Mini^3 but is instead powered by USB alongside the BantamDAC in the Hammond 1455C801. Each USB port offers 2.25W at full power, and the circuit is powered by a 1W DC/DC converter. The 12V boosted voltage is filtered by a CLC filter and regulated by a 10V LDO regulator. The layout is a little different from the Mini^3's, but it borrows very much as many of you will notice.

Let me continue by explaining my intentions for this new project. My Toshiba Satellite A100 notebook was my main source of music, and I used it every day at school for almost half of every day. About a year ago, the physical volume knob snapped off. To replace my newly lost music, I wanted a solution that was extra portable and convenient. To me, this meant no battery, as few wires as I could get away with, and of course, high quality audio. I started work on a new DAC+amp board, but just as I found the EAGLE library for the PCM2702, the BantamDAC was gaining momentum and became a very popular choice for USB audio. At this point, I had an epiphany: I could use the BantamDAC, which so happened to fit the Hammond, and could just design the amplifier part. Fast forward a couple more months, I asked amb if I could use his Mini^3 amplifier circuit. He agreed if I would give credit to him and if the circuit performed as well as the Mini^3 does. Though at this point I'm not sure this thing will even work, I feel it's all necessary to share in the spirit of responsible invention. If things don't work out, maybe someone will learn from my folly and finish the race for me. I have learned lots from this forum, so this is my attempt at giving back. If all goes well, people will be able to enjoy a truly portable and compact hi-fi headphone amplifer.[/size]

Amplifier

The amplifier schematic is identical to the Mini^3. The layout varies slightly because of the method of signal input. With one less 3.5mm jack on the board, that space can house the L/R section to increase room for the power supply at the rear of the board. This allows for physically and electrically larger capacitors and a smoother power supply. At the top of the board we find a power plane for the negative rail, which ought to provide better cooling for the EPAD underneath the AD8397 as well as a low-impedance path for the return currents.

Power Supply

The 5V 500mA power from the USB is converted to 12V. The TI DCP010512P DC/DC converter is rated for 1W, so there is an output current of 83mA. The DC/DC converter introduces noise at 800 kHz, so a CLC filter with a corner frequency of 580Hz cleans up most of the noise. That voltage is regulated to 10V with the TI TL750L10. The higher supply voltage also gives the amplifier higher voltage swing than the Mini^3.

One tricky part of using USB power is grounding. The designer must figure out how to separate the signal ground with the USB ground if a virtual grounded amplifier is to be used. One way of solving the problem of the grounding is by using an isolated DC/DC converter. Combining the amplifier with the BantamDAC means connecting signal ground to V-, so with an isolated converter, the output can then float to whatever value relative to signal ground, that is, +/- 5V.

Resources:
USB powered AlienDAC with CMoy
http://www.head-fi.org/forums/f6/usb...27/index2.html

USB-powered amp/ps
http://www.head-fi.org/forums/f6/usb...amp-ps-215585/

USB powered HIFI headphone amp/DAC
USB powered HIFI headphone amp/DAC - diyAudio

 

[joneeboi]

Schematic v1.23

Layout v1.23

 

[rds]

Quote:

rds wanted to power his PIMETA, but there is the tricky issue of the virtual ground versus "true earth" ground.

It's not a problem actually. The dc-dc converter is isolated.

Quote:

The biggest hurdle here, I feel, is power supply.

Well, good news - I'm going to build the supply separately on breadboard and break into my school's lab to measure it (and maybe tweak it).

I will post the circuit diagram and parts list as well as the measurements.
It will be a regulated 12V galvanic isolated dc supply that can supply 150mA from usb 2.0.
Then people can take that and use it to power whatever you want. No need to worry about virtual grounds or any related nonsense.

I'll have that posted in 2 weeks.

The cost to build it will be about $16 USD.

If there's interest and it measures well I might even fab some pcbs, but no guarantees on that. Pcbs for dc-dc converters need to be very carefully designed.

Anyways, it'll be easy to build.

EDIT I'll try a capacitance multiplier, as suggested by 00940, too. I might get lucky and be able to get the voltage drop low enough to put it between the dc-dc converter and 12V ldo. However, my feeling is that the capacitance multiplier won't be any use at the 800kHz point where I'm trying to get rid of the ripple.

 

[joneeboi]

So you're going with 12V then, eh? Which chips did you end up using? I was thinking of getting a run of boards done since I've always been curious about how that all works. Either I can get the PS and amplifier done separately for whoever wants to power whatever, or I can throw it all onto one board. Maybe I'll just stick to the one board for now. I'll keep my eyes peeled for that two week mark. Thanks for the heads up, rds.

 

[rds]

Quote:

Which chips did you end up using?

Parts information here

I'm going to put a pi filter between the converter and ldo. This will use 1uF cermics to effectively filter out that high switcher frequency. Since it is so high I don't need a big inductor to do this.
I'll have another pi filter after the ldo. If the switching ripple is nearly gone then I'll use it with big electrolytics to filter low frequency. I'll have to see what I measure. I'll probably just optimize it for wherever there is the most ripple.
But it's looking very good by a few ideal calculations. I see ti recommends pi filters for their switchers too, so I'm optimistic.

 

[millwood]

if you use a r2r opamp you can swing fairly close to the rail so that will get you 2vp on the load, conservatively estimating.

that will produce over 100mw on a 16ohm headphone. that kind of power is typically sufficient for most listening but if you want more, you will have to swing the voltage.

the safetiest route is a switching mode chip that produces negative supply from a positive rail - plenty of them in various datasheets if you are interested.

alternatively, you can use one single step-up converter and stay single rail - you will have to use capacitors.

if you don't want capacitors, you can use also fly-back based dc-dc converters. the dcv family of chips from TI would be a good starting point.

 

[millwood]

the last alternative would be use one boost up converter and then great virtual ground with it.

Personally, I would use a voltage intertor to create the negative rail.

 

[rds]

Quote:

Personally, I would use a voltage intertor to create the negative rail.

I definitely wouldn't use such a supply for audio.

 

[dsavitsk]

Quote:

Originally Posted by joneeboi /img/forum/go_quote.gif [*]USB-powered[*]Fit into the Hammond 1455C80x[*]3-channel CMoy amplifier[*]Low-cost[*]Relatively easy to build[/LIST]

I would go 2 channel as, frankly, the third doesn't get you much but it does draw a bunch more current.

For an easier boost converter, take a look at TI's DCP020509 and DCP020507. A little more expensive than some other options, but they have DIP pins, and a pair can be used to make a +/- supply.

 

[Mellow Mushroom]

why not use battery power instead?

 

[rds]

Quote:

why not use battery power instead?

Because the design requirement specifies no batteries.

 

[Mellow Mushroom]

Quote:

Originally Posted by rds /img/forum/go_quote.gif Because the design requirement specifies no batteries.

But but... but his laptop uses bat batteries.

 

[error401]

Quote:

Originally Posted by rds /img/forum/go_quote.gif If there's interest and it measures well I might even fab some pcbs, but no guarantees on that. Pcbs for dc-dc converters need to be very carefully designed.

Don't think that'll be a big deal. All the high speed stuff is contained on the IC die, so you don't have to worry so much about it. Make sure the decoupling is good and as specified and it will work well.

Quote:

Originally Posted by rds /img/forum/go_quote.gif EDIT I'll try a capacitance multiplier, as suggested by 00940, too. I might get lucky and be able to get the voltage drop low enough to put it between the dc-dc converter and 12V ldo. However, my feeling is that the capacitance multiplier won't be any use at the 800kHz point where I'm trying to get rid of the ripple.

Many LDOs are fast, but many aren't, and it's not always shown in the datasheet. Mostly ripple rejection is specified at 120Hz or 60Hz, and if there's not a graph out to at least 1MHz, you can't really speculate on performance in that region. A cap multiplier can be made very fast and very cheap. Consider the attached that I knocked together in about 5 minutes and did virtually no tuning with. The response is fairly load dependent, it gets better with bigger loads. The sim is at ~100mA; it stays under -50dB @ 800KHz to about 55mA and gets better with increasing loads. I haven't simmed the load response, but it should be fine, and most of the load is probably going to be static anyway. Probably will need an output cap. Drops about 1.5V.

Edit: Why does head-fi take my 9kb PNG files and transcode them to smaller, much junkier looking JPGs?!

 

[rds]

I'm not assuming anything about the psrr of the ldo. I'd expect it to be quite poor at high freq.
The thing is a capacitance multiple can't do anything about the esr (afaik). And if you look at the dc-dc converter datasheet you see that low esr has a far greater effect on reducing ripple than high capacitance.
My plan for getting rid of the ripple is a pi filter between the dc-dc and ldo, and possibly a second pi filter after the ldo. I looked at some t.i. reports and this is actually what they recommend where is is not cost or space prohibited. The reason being that even a small L will cause huge attenuation at 800kHz.
Also worth mentioning is the L is about $2 - so not too bad when you consider the dc-dc is $15. Oh, and the low esr ceramics are about $1 each