Member Since: July 6, 2007

Country: United States

  • s/11/7/ since the LCD controller can be put into "4 bit" mode where only the upper nibble is needed to send/receive data. Indeed, in a pinch it's even possible to get by with just five (D4-D7 and RS) if R/W is tied low and one depends on hard-coded timeouts that are "long enough" so that one can dispense with checking the state of E.

  • Easy peasy. A saturated solution of many common salts will reach an equilibrium with the surrounding air. Regular NaCl and distilled water is the magic 75%. You need a "slushy" mix with distilled water and pure salt. There should be an excess of the solid salt to ensure that it is a saturated solution.

    Magnesium chloride is commonly used for the 33% measurement in a two-point calibration. Somewhat harder to find. Don't use the rough stuff sold for de-icing; not pure enough.

    Table at http://www.omega.com/temperature/Z/pdf/z103.pdf

  • It's not an STM32F103RCBT6. The STM32F103Rx processors are used in lots of dev boards so the 103R "finger macro" isn't too surprising. ;-)

    Actual chip is the STM32F103 C (48 pins) B (128 K) T (LQFT) 6 (-40 to +85 C).

  • 3.3 is here to stay. Fortunately, 3.3 5 interoperation is often transparent since virtually all 3.3 V microcontrollers will accept 5 V inputs on some or all I/Os and can drive TTL-compatible loads just fine.

    As always, check the data sheet. The STM32F10x minimum guaranteed output level for a logic high into a TTL load is 2.4 V, and the guaranteed minimum input threshold for older TTL and modern HCT TTL-like parts is 2.0 V, so it's good to go, modulo noise margin etc.

    Not all STM32F10x inputs are 5 V compatible so again, check the data sheet. A voltage divider may be all that's necessary.

  • One note on the 2/20/200/... range series.
    Manufacturing cost is probably one issue, as noted. The other contribution, however, is that you get readings that are "three significant figures" across the scale.
    If you have a slide rule handy (or, if not take a look at some log-log or semi-log graph paper) notice that the spacing around, say, 1.234 is about the same as around 0.987. It's a reasonable practice, when counting significant figures, to disregard the leading digit if it's a one for just that reason.
    See also discussions of Benford's law for how this phenomenon shows up in other places.

  • Just to follow-up on this ...
    Tagging a couple 6.8Ks to the SCL and SDA lines helps a lot to clean up the appearance of the I2C bus. There's plenty of room to superglue a couple of 0805s to the soldermask and then add some "field change" 30 gauge wires to the processor pins and also to 3.3 V.
    The issue with the magnetometer seems to be that it really, really wants a full power cycle or else it can get hung at a fixed value (32 in my case; don't know if that's universal). It wasn't an I2C bus problem but I'm happier now that the bus looks right.

  • Very unhappy I2C bus, at least on this sample of one that I have here.
    It looks like the board is depending on the 50K pull-ups that are internal to the magnetometer and that is way too much to run a 200 Kbps I2C bus. I'm seeing a rise time (standard 10%-90%) of more than 2 usec, twice the maximum for a even a "standard mode" 100 Kbps I2C bus.
    The accelerometer (which shares the I2C bus) seems to be okay with it but the magnetometer isn't talking. Maybe it's dead but I suspect that tagging the SCL and SDA lines with 7.5K or 6.8K will do the trick.

  • Yes, it is. Well, only the crystal is removable; the socket is soldered to the board. ;-)
    It's the usual HC49/U or HC49/US thru-hole form-factor.

  • Those are the A/D readings assuming that the vane sensor is on the lower (grounded) side of a voltage divider with one of the resistors at the head of the columns on the top (Vcc) side, and that the A/D resolution is 8 bits.
    E.g., for a north wind (33K ohms) and a 10K dropping resistor, 255 * 33K/(10K + 33K) = 196 counts. There's certainly no requirement to use any of those resistor values but they're quite reasonable.
    It's not the clearest data sheet I've ever seen. Not the worst, either. ;-)

  • The wind vane outputs are discrete, not continuous, with eight 45-degree sectors more or less centered on the compass rose. Different resistors are switched in-line for each sector. One caveat: the boundaries are not crisp and you may see an open or an overlap (with two resistors in parallel) for a few degrees around each sector edge.
    Values in K-ohms:
    N 33
    NE 8.2
    E 1
    SE 2.2
    S 3.9
    SW 16
    W 120
    NW 68

No public wish lists :(