[Empeg-technical] Re: Possible alternate source for VFD

Hugo Fiennes altman at empeg.com
Wed Apr 9 13:41:00 PDT 2003


The IDE isn't very fast on the SA1100 as the cycles are fairly long. The SA1110 helps a bit (hence why the Rio Central has faster IDE than the carplayer).

Given a PXA255 like I have on my desk here, I can get more like 8mbytes/sec via PIO simply due to better timing options in the chips IO block. Even more with DMA  :-)

Ethernet can be done by the SMSC chip without many problems, but there aren't "lots of" non-PCI 100mbit chips; the SMSC one was the only one I found, last time I looked.

The best solution would likely be something like the Cirrus 9312, if it ever comes out. It has (onboard) IDE-DMA, 100mbit ethernet, USB host, etc. Again, I have one of these on my desk, but it's an old rev and was very buggy (and it's already 2 years late) - it also may not drive the VFD without some hackery.

A PXA would be nice for outright speed (400 vs 200MHz of the Cirrus), and ARM9E extensions (for audio algorithms), but you then need to add ethernet < a USB host chip.

The big issues are:

- 4-channel audio out. Use a DSP or just do it all in software with 2 stereo DACs? If you go this route, what will do the FM MPX/RDS decode as well as the DSP? Going the other way, I don't know how many software changes would be required to get the latest (ie, only available) philips DSP working.

- Software support. It's likely that we'll aim for a 2.4 kernel in 3.0, to ease things like ext3, but there's still going to be far more work involved than could get done in spare time.

- Cost. Both CPUs are BGA, so "kit form" is going to be out. It's likely to be spendy.

Hugo  




More information about the Empeg-technical mailing list