audiophile diy DAP? - Page 2

Making something small with decent battery life that could play FLACs (for example) would be extremely difficult without some uber computer engineering skills. Especially if you want to interface IDE or SATA to it.

I suppose hacking something something like an NSLU2 wouldnt be impossible. You could simply hook up a 2.5" USB hard drive to it, and interface a serial LCD to the serial port. The CPU, 266mhz Xscale, should have plently of power to decode FLACs, but you'd have to be an embedded linux guru to figure out how to get digital audio out of the thing (from one of the USB ports, perhaps). Even so, there would still be the problem of how to run the thing on batteries, and it would still not exactly be small.

Edit: I take it back, apparently its quite easy to interface a USB DAC to the NSLU2. http://www.nslu2-linux.org/wiki/HowTo/SlugAsAudioPlayer
I suppose interfacing something like a PCM2702 to it would be trivial then. The only remaining issue would be battery power. http://www.nslu2-linux.org/wiki/HowT...eryPoweredSlug

TI has ATA interface chips, DSPs specially designed for mp3 players, DACs, and application notes on how to wire that stuff together

other IC fabs surely have similar product lines

http://store.makezine.com/ProductDet...tCode=MKMP3KIT

May be one way to achieve a good player is to use a ITX Pc, the nano itx motherboards is 12cm x 12cm and has everything you need. I am not sure they are fanless but it can use external PSU. If you want to go smaller you can go with the Pico Itx it's 10cm x 7 cm.
Via who promote the Itx format is known for their ship in the av710 soundcard. The price of Nano, mini and pico Itx decrease regularily, So wait and see.

Another way is to use A board dedicated to music like this one or this other one. sorry the second one is in french. But it is a pure Diy, they sell Pcb, take a look to the ships, I don't think they will be easy to solder.

I will prefer the first solution, it's the more flexible one.

The reason it is so hard is obviously the components used are typically proprietary and are only manufactured for the company making it. Like the LCD screen for example, and the really, really, really complicated PCB's that have traces so small its scary.

thanks to the trend towards sticking everything on one chip, most of the ata/SD interface/etc external chips aren't really necessary--a lot of microcontrollers already have these built in.

If there's people interested in wallwart powered, somewhat large DAPs, well that's actually not too difficult to make. However, if we want a portable device, it's a lot more difficult. I've been looking into DAP design a lot these days, and it appears that the main difficulty for us HF DIYers wouldn't be so much the engineering behind it but the uber-unobtainium required. I've looked at several DAP schematics from the rockbox project, and honestly the hardware for most DAPs is virtually the same--a DAC/headphone driver IC, some power management ICs, display, RAM, and a microcontroller which sometimes has the rest built-in. They differ in the displays used, the casing design, or the interface design (i.e. capacitative touch sensor on the iPod), and the exact ICs that are used. The most difficult part of hardware design for us is most likely the PCB routing and finding out where to get the casing manufactured.

More difficult is sourcing the components which will allow us to make a low-power and small DAP. Unlike our headphone amplifiers, reducing power consumption is something that can really only be done on the IC manufacturer/designer side. If we want our own DAP to compete in terms of performance/size/battery life with the latest generation of DAPs (iPod nano, etc) we would need access to the newest/most efficient microcontrollers, RAM chips, flash ICs, etc. . . which is mostly uber-unobtainium for DIYers. If we just want something like the old iriver H3xx, we're faced with another problem: since we're looking for audiophile quality, and audiophile = uber-inefficiency, chances are, our design would be even less efficient/larger than the old DAPs which lasted only 8 hours or less. I'm pretty certain that most of our current headphone amp designs/DAC designs already use more power than an entire portable DAP.

The LCD screen for the portable DAPs was indeed the most difficult thing to find when I was first researching the feasibility of this project. I finally got referred to sparkfun.com, which sold small color LCDs for relatively reasonable prices (I think mouser started stocking small color LCDs, too, but at $100+ per unit). Now they even managed to get a hold of genuine PSP LCDs at what I'd say is a pretty reasonable price--they're absolutely the best source for diy uber-unobtainium I've seen on the Internet. Hopefully, should we try going forward with a DIY portable dap, they might be able to help in finding the rest of the uber-unobtainium.

What I think might be able to allow us to make a truly portable, competitive DAP is with a lot of CS trickery. If we've got any EE+CS geniuses around here, I've got a couple of (possibly impossible) challenges for you that could really make a truly revolutionary DAP:
a) get a FLAC/MP3/shorten (shorten seems to use the least processing power, at least on the x86, iirc) decoder to run realtime on a 16-bit MSP430 chip, possibly running in parallel (these things can run literally ~300uA/2.2V at 1Mhz; at higher speeds it's still pretty darn low power)
b) use a more complex microcontroller and instead of mathematically decoding everything, see if you can pull it off using lots of lookup tables. This theoretically could significantly reduce the clock speed and power required. There's a lot of problems with this, including ram timing, what kind of memory to use, etc. . . but philosophically it just kind of appalls me to think that our computers have to recalculate the same basic functions over and over again

hm... I really like making looooooooooong technical posts, don't I?

Sure, with a team of CS, CE, and EE people putting thousands of man-hours of work into this, it could become a reality. But, I think this is outside the scope of a DIY project.

It's sort of strange really, that there are such massive open source software projects around, but open source hardware has yet to really materialize. I guess Arduino and Chumby are pioneers in that field, but even those are somewhat less complicated than this. Arduino is just an AVR platform, and Chumby doesn't have to worry about being power efficient.

If you do want to get a full scale open source hardware project together though, I'm up for helping out. I'm currently studying Electrical and Computer Engineering, so I've got some background in the area.

Well another option would be a "transportable" dap- by that I mean something not so small as to fit into your pocket, but perhaps something that would fit into a bag or backpack... That would make things a bit easier right?

As far as the LCD goes, I know that one DIY MP3 player site sais it uses screens from an old model nokia cell phone- you could probably buy used/broken units on ebay for cheap

Interesting. I'm going to subscribe to this thread to see what happens.

The codecs are the biggest challenge. There are plenty of very low power dedicated DSP products for decoding MP3, WMA, etc., but none that I'm aware of that handle lossless codecs such as FLAC.

codecs aren't at all a problem if we can stick rockbox on it, and since I'm going to be running the same cpu as the h340, it should be quite possible.

The main difficulty in my head at the moment is a bdm programmer for the coldfire microcontrollers (used on the h340/compatible with rockbox) that doesn't cost $250+; honestly I don't think I'll be programming on that platform in the future, and I'm not ready to drop that much for something I won't use again. Anyone here played with fpgas/cplds/programmable logic before? I found just one schematic for a bdm programmer for this chip, but it uses a programmable logic IC which I have no idea how to use/debug/program. There's also bdm programmers for other platforms, like the 68k--would those also work with the coldfire microcontrollers?
I know most people on head-fi probalby wouldn't have much experience with this digital stuff, since we mainly deal with analog, but I guess it's worth a try with some questions anyway.

And back to analog! I've never really looked into audiophile DAC design, so I'm just wondering exactly what makes a DAC audiophile quality? I've heard of jitter, but exactly how do we minimize it? Are there any other parameters that should be considered when designing a DAC, and what DAC ICs should I use?

Forget all about codecs and fancy processors. Just run .wav files of a 2.5" IDE harddrive. They are easy to decode and even a 40 GB harddrive could hold over 400 songs at 24 bit/192 kHz. Throw in a CS4398 or WM8740 DAC and your set.

AFAIK, there are no FLAC DSPs, or low power FPGAs, for that matter. That's why it would be easier just to use software decoding, and have a reasonably beefy CPU. If it's not going to be small anyway, that should be fine.

Anyway, I looked at my friend's NSLU2 again, and it's much smaller than I remember. With a bit of effort, it would be possible to fit an NSLU2, a small LCD, a 1.8" hard drive, and a small DAC inside an original Gameboy case. a battery wouldn't fit inside though.

Anyway, I think the best bet is to go for something Xscale, simply for ease of programming (and because several hardware platforms already exist, making this project, well... realistic).

May be someone can take a look inside a high end Dap and tell us what are the ships used used inside ?