Hi Here’s the issue about “big powerful 32 bit ARM processors”….
At the chip level (as in no board, just the bare part to solder down) their are many parts below $1 and some below $0.50 in reasonable quantity (say 10K). It’s hard to find a useful MCU of any sort below $0.25, so the “premium” for an ARM is not very large. Depending on what you need for i/o, the ARM may / may not be lower priced than the alternative. In a *lot* of cases it is the low price option. There are many flavors of ARM and many competing parts, so doing a full matrix to compare them all is a massive undertaking. Indeed a Coretex M0 with 32K of flash is not the same thing as an M4 or something bigger still. There are a lot of flavors of ARM. At the semiconductor level, none of the manufacturers are going to state it clearly, but it’s becoming more and more clear. The support for (and enhancement of) chips that are not based on ARM is going to dry up. It’s happening already if you watch the little details. It will be happening more and more as time marches on. It will be *very* tough for an MCU over about $1.00 a chip (at say 10K pcs) to compete with the ARM parts going forward outside niche markets (like DSP). No, it’s not going to happen overnight. It will take a number of years for things to work out. ——— So why do a GPSDO on an ARM - They have staying power. The bang for the buck is there. There are a lot of cheap boards out there with them already set up and running on them. Nobody big is going back to leaded parts. Semiconductor guys follow the big users. The future of “home project” work with CPU’s is to use a cheap board rather than a chip. The prices of the boards will just keep coming down, they are not going to go up…. I can’t buy the parts on a $10 board for $30, let alone put them all together. Pick a board and go with it. Put your *stuff* on a second board and plug them together. Cheaper, faster, easier than DIY. You get USB / Ethernet / whatever thrown in. With the right board (chip) you get flexible clocks and PLL’s that make the whole process a lot easier. What’s not to like? Well yes, you need to dig into a tool chain. They either are free, or are going to be free fairly soon now. There are even web based tools to do the code stuff (mbed). You can’t quite write it for one chip and drop it on another one blindly. You do get locked (a bit) into your chip choice. All in all, far more advantages than disadvantages. Bob On Apr 10, 2014, at 8:11 PM, Chris Albertson <[email protected]> wrote: > I'm thinking about good reasons to build a GPSDO using something as big and > powerful as a 32-bit ARM processor. I think the reason is that you are > not really building a GPSDO but some other device that just happens to have > a GPSDO inside of it. > > For example you want to build a laser range finder and you need to measure > time of flight delay. You'd need a very good clock and while you are at it > why not discipline the clock with GPS. I could think of some radio > experiments where I would want pairs of receivers with their local > oscillators running in phase but many miles apart, so I'd build a GPSDO > into the radio. The ARM would support running the GPSDO, the bigger > application and also remote access over the network or Internet. > > Today people mostly will build a stand alone GPSDO in a dedicated box and > then connect the 10MHZ output to whatever is needed but now as we have > seen, you can build a GPSDO completely in software, if your project already > has a computer then you can run a GPSDO inside an interrupt handler as a > background task. Adding GPSDO functionality to an existing product is > almost trivially simple, just $2 in parts and some software if you already > have a CPU and OCXO as part of your system. > > Placing the GPSDO inside the product means the gpsdo can run at a frequency > that is more useful and needs no conversion. So you can have the GPS > control a 23 Mhz crystal if that is what your laser rangefinder needs. Now > that the cost of a GPSDO has fallen to $3 you can build them into > "everything". It no longer needs to be a shared device. > _______________________________________________ > time-nuts mailing list -- [email protected] > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
