How to carve out 3 1-bit ports from a 4-bit port?

[segher]

Yes but you can switch between Input and output, albeit all bits at once

That's right; but you should make sure it is okay with the device connected to the other
if you drive a pin that it can drive as well; and the other way around, make sure it is
okay to float a pin that the other side uses for input.

making some protocols possible, like perhaps slave I2C?

For IIC, you need to do both input and output anyway: output when you are sending
a 0, and input when you are sending a 1 (or receiving).

[Folknology]

For IIC, you need to do both input and output anyway

Sure.

output when you are sending
a 0, and input when you are sending a 1 (or receiving).

Love that ruthless reduction of open collector operation of IIC.

regards
Al

[landon]

Thanks for the discussion, everyone.

This lack of 1-bit ports is forcing me off the XK-1 sooner than I expected.

Al, I agree with your assessment on the architecture improvements. It would be splendid to treat any pin as a one-bit port even if there was a fixed number of them less than the total port/pin count.

My only other huge wish on XMOS improvements is more RAM. One of my 1-bit needs because of lack of RAM is to talk to a 4-wire F-RAM and an SD for archival.

Despite various limitations, still loving learning about and using XMOS,

Landon

[landon]

BTW - I think my strategy, were I to stay on XK-1 would be:

reserve all the 1-bit ports that are broken out for things like clock, MISO, and MOSI. Then put all the chip select lines bit one of each of the 4 bit ports...that would cover 2-3 SPI devices depending upon whether it's a MISO and MOSI requirement.

This way I could use all the nice XMOS clocked ports, and clocked i/o on one-bit ports and leave the simpler select parts for gpio on 4-bit ports. This approach will stretch the 1-bit ports further before having to move off XK-1.

Landon

[bsmithyman]

I think a lot of problems would be solved if the SPI lines used for the flash were broken out somewhere on the XK-1. They're already set up for SPI anyway, so it just requires one additional pin per device for the CS signal.

My only other huge wish on XMOS improvements is more RAM. One of my 1-bit needs because of lack of RAM is to talk to a 4-wire F-RAM and an SD for archival.

By any chance are you using one of the Ramtron F-RAM chips in SO-8 package? They're pretty speedy, and non-volatile as well; not much pricier than an SPI RAM anyway, and they run up to 40 MHz. I actually wondered if they could be used in place of the SPI flash, but I think the lack of a Page Program instruction would make it difficult.

[landon]

Yes re: Ramtron F-RAM - SOIC-8:

Mouser Part #: 877-FM25V10-G
Manufacturer Part #: FM25V10-G

I see Mouser is out of them and they're quoting a 14-week lead-time. I bought 4 of them at the time, so hopefully, that'll get me through prototyping. Where are some other good places to source the Ramtron devices?

I am actually fine with the XK-1 not breaking out its SPI Flash lines....I would prefer to keep my peripheral SPI bus(es) separate from my firmware.

I also thought I'd experiment with a serial SRAM but the SRAM sizes are not that large...I have a 256Kb On Semi SRAM (Mouser #: 863-N25S830HAS22I Mfr. #: N25S830HAS22I). I would love to find a larger than 256Kb serial SRAM device. If you know of one, definitely point it out to me.

Mouser is terrible about distinguishing between a KB and a Kb. Boggles me. The On Semi SRAM for example is titled 256KB but is a 256Kb device.

Landon

[bsmithyman]

I sampled a couple of the 2 Mbit chips from Ramtron (FM25H20), but they're pretty pricey (i.e. $20/ea). Not that it's necessarily unreasonable, it's just what FRAM costs. Also, they draw more than 3 times the current of the FM25V* chips, if that's a factor. I usually buy from Digikey, and I find they're pretty good about the bit/byte sizes, but they don't seem to have Ramtron parts at all. The nice thing about the FRAM is that they're fast and they've got a preposterously long lifetime for non-volatile chips (like 10 years at 100% I/O load). For some reason I feel like I saw a product bulletin that they're working on 4 Mbit and 8 Mbit chips, but that they aren't in full production yet.
Looks like Future Electronics also stocks Ramtron parts, but they're mostly at zero quantity at the moment.
Edit: Maybe http://www.alliedelec.com/?