audiophile diy DAP?

[j4cbo]

I don't think Xscale is available in anything non-BGA, which would make DIY assembly more or less impossible.

The NXP (formerly Philips) LPC2368 is a nice processor: 72 MHz ARM7, half a meg of Flash, built-in I2S, built-in SD (8 gigabyte SDHC cards are cheap now), etc. It only has 32k of internal SRAM on the bus, though, which may be insufficient.

[threepointone]

Yeah, people already have. That's what I'm basing my design on--the iriver H340. I was pretty much inspired by everything by rockbox' work on disassembling and documenting all the hardware in the DAPs; I realized that there the differences among most DAPs were minimal, and that as long as if we had the software to run everything (rockbox) the hardware wouldn't be that difficult to design (of course, sourcing all the parts and miniaturizing/reducing size would not be as easy). For example, look at the Iaudio X5/M3 and Iriver H3xx/H1xx; the basic components used in those players are virtually identical in function, and those players actually use the same processors as well. The other rockbox players, including the iPod, Gigabeat, and H10 also use pretty much the same portalplayer ICs (which are not accessible to us).
Hardware info is available via http://www.rockbox.org/twiki/bin/view/Main/WebHome. Click on the player you're looking at under "Ports" then look for the link labelled "[player name]hardwarecomponents" or the schematics, if available (didn't want to post 15 different links)

Anyway, heres the basic idea of what i'm going to be working on:
a) first I'll probably prototype a (mostly) non-portable wall-powered unit based on the MCF5250, the microcontroller used in the H3x0. It'd really just be a quick prototype/feasibility check project.
b) then I'd start on a portable version. I might still use the MCF5250 first, but I think the base power consumption of that chip might be too high. I might look for a lower power ARM chip.
c) completely radical design. might pick some exotic ultra-low-power architecture or see if I can get some of the dedicated DAP chips. Might run a custom firmware instead of rockbox.

of course, everything would be designed with audiophile quality in mind.

The primary things that are preventing me from going forward at the moment are:
a) need a way of programming a coldfire microcontroller. The standard way is to use a BDM programmer, but those cost $250+ minimum and I don't see myself using coldfire ICs after this project so I'd rather make one myself. The only diy coldfire bdm pod I've seen requires programming cplds, which I don't know how to use either. There are some DIY BDMs for other kinds of microcontrollers, but I don't know if they'd work for the coldfire platform. The coldfire microcontroller also has a JTAG interface, but I'm still not entirely sure if I could use that for programming, as it is usually used for testing during production.
b) prototyping PCBs. It's not really a problem specific for this project, but currently I've got another project on my mind for making prototype PCBs which might go first to ease the wait time for the DAP project
c) MCF5250 IC. Waiting for samples--hopefully freescale accepts my request (I put legitimate info, of course). If they don't, I'm screwed because the only other place that has MCF5250 (or scf? i still can't get this straight) ICs available has a pretty large minimum order quantity, which also poses some problems if any of these designs really work and become published. May need some group buys and/or redistribute parts

[spongezone]

Originally Posted by j4cbo

I don't think Xscale is available in anything non-BGA, which would make DIY assembly more or less impossible.

The NXP (formerly Philips) LPC2368 is a nice processor: 72 MHz ARM7, half a meg of Flash, built-in I2S, built-in SD (8 gigabyte SDHC cards are cheap now), etc. It only has 32k of internal SRAM on the bus, though, which may be insufficient.

My point was that there are many Xscale boards readily available at reasonable prices, making a DIY project realistic. And more so, it would make it truely DIY, rather than a, one person buys expensive equipment, manufactures boards, and sells them to people who want to "DIY" it.

Just a few examples are NSLU2, Gumstix, and Zipit.

[utilisateur]

*subscribed

and i've got a link to check out
NSLU as audioplayer

[spongezone]

Originally Posted by utilisateur

*subscribed

and i've got a link to check out
NSLU as audioplayer

Not bad. Looks like FLAC is only 17% CPU usage for a 266mhz Xscale.

[threepointone]

If we're still talking about making a portable DAP, we shouldn't be looking at clock speeds at all--rather, we should be looking at how little power it'll use at lower clock speeds. Sure, the iPod ARM processor might be able to run at 75Mhz or higher, but most of the time it probably runs at lower speeds in order to save power. Since the XScale is rated for such high speeds, though it may be low power consumption for its performance, I wouldn't be so sure that it'd be the best choice for lower speeds. I'm not very familiar with the XScale series, so I'm not sure exactly which processor I should look at to evaluate, but I would think that it also costs more than alternatives.

Otherwise, if we're just talking about a desktop DAP, I really don't think there's a problem with any of the chips suggested here--the ARM series is used by many (if not most) modern DAPs. The XScale series is actually based on the ARM core, according to wikipedia. Additionally, the ARM (or at least the portalplayer flavor) is already running rockbox, which is the open-source replacement firmware for the ipod/iriver hxx(x)/iaudio/toshiba gigabeat/etc that can play FLACs. Personally, I'm looking at the coldfire SCF5250 series right now since it's the most obtainable and most usable (available in non-BGA surface mount) microcontrollers out of all those supported by rockbox. I would expect that we'd have to modify <50 lines of code; in fact, we'd probably be mostly deleting modules from rockbox, since we wouldn't need any of the power management routines.

[threepointone]

Originally Posted by j4cbo

It only has 32k of internal SRAM on the bus, though, which may be insufficient.

Don't worry about the internal RAM; virtually all DAPs (or any of the more complex digital equipment) will have an additional external RAM chip.

[j4cbo]

Originally Posted by threepointone

Don't worry about the internal RAM; virtually all DAPs (or any of the more complex digital equipment) will have an additional external RAM chip.

Necessary for buffering in a hard drive player, but not so much if it's Flash-based. External RAM is certainly an option but adds quite a bit to the design difficulty.

I suspect you're underestimating the effort required to integrate all the hardware and software here. Using a chip that Rockbox has already been run on doesn't really provide any advantage, as it'll still be essentially a new hardware platform (everything but the uC, user interface available, etc.); why limit yourself?

[bubbamc119]

Great idea. 24bit playback should be considered.

[Cauhtemoc]

Originally Posted by j4cbo

Necessary for buffering in a hard drive player, but not so much if it's Flash-based. External RAM is certainly an option but adds quite a bit to the design difficulty.

32 kB is enough to store some 30 ms of audio at 24 bit/192 kHz. Even the slowest of harddrives doesn't have a random access time this long.

[bubbamc119]

I'm going to subscribe to this thread. I finish my electronics/comms degree in a few months and would be glad to help get this off the ground.

[labmat]

I would love to see something like this get off the ground; and I would love to help wherever needed. I'm a Mechanical Designer by day and have access to 3D CAD software and can design any mechanical parts needed.

[Danielj]

Wav of course is mandatory and flac would be nice. If wav only, then of course the case would have to support much larger drive. Either way, one thing I think would be essentail is hardrive swapping or more importantly, access to a 2nd HD via usb. I have noticed in reviews of these new players running off the sigmatel chip, is that they are having problems accessing a HD via usb...seems to be a firmware problem.
Perhaps in our player; a physical switch to select either internal or external HD prior to bootup. Why would the chip care which one is being used?

[threepointone]

interfacing an HD via USB is completely different than interfacing with an directly with an HD (ATA protocol). however, if we really want it as a feature, I know FTDI makes a chip that can handle USB mass storage devices; we can add it at the expense of complexity/space. Having the USB host on the CPU is a nicer way to do it, of course, but USB host features seem to be not as common and probably require more software without a very high level dedicated IC for it

[Danielj]

Well, then...flac with room in the case for a 120-160 gig lpatop size HD.

Other thoughts The dac should be given high priority in the design; if we use the same sort that is in the portable players available...what's the point, just by an Iriver. I only know what I read about dacs; monica 2 and alien dac. the Spitfire by firestone has good reviews and they mention the components used. Also many people say good things about the dac used in the emu systems external soundcards.

Also, most portable daps do not have a true line out, they just lock the amped signal at a certain gain. If our project has a headphone jack, then again perhaps a switch for lineout that bypassed all the amp stuff.