Group Actions
Engagement
  • Author Author: shabaz
  • Date Created: Date Created
  • Views 1657 views
  • Likes 11 likes
  • Comments 12 comments
Related
Recommended

Soldering Tiny Temperature Sensors

shabaz

I've been experimenting with temperature sensors, and really liked Sensirion products, since they are pretty accurate and reasonably low cost, and have a sensible board-level interface, I2C. However, they are quite small!

The good thing, is that they are not too difficult to solder by hand. This ultra-short blog post shows how I went about it.

The part is 1.5 mm on its longest edge (i.e. it is smaller than the thickness of a normal PCB). The photo shows a 0.5 mm pencil for a size comparison.

So, first of all a decent pair of tweezers is needed, otherwise the part will go flying. Squared-end tweezers provide a lot of grip on the part, and it is a reliable way to manipulate the part. Good eyewear or a magnifier is needed to notice the tiny notch on the underside, which indicates pin 1 (there is a marking on the top side too). I used a mobile phone camera for this snapshot, and it was good enough to be sure of the part orientation.

image

I could have used a hot air tool, but I wanted to stick with a soldering iron for this exercise. First off, I applied flux paste to the four pads on the circuit board, which happened to have a small amount of solder already on it as manufactured (it's known as a HASL finish on the board). HASL isn't great, but it's cheap.

Another thing to pay attention to, is the component footprint. Unless you've got excellent eyesight or a microscope, you're going to be relying a lot on the silkscreen outline to align the part to the PCB. Therefore, it's quite important to have a decent silkscreen that is of the correct dimensions to match the part. In the screenshot below, since the part is 1.5 mm square, I made sure that the silkscreen was that size too (with a fraction of a millimeter exceeding the package size, so that I could place the part and still be able to see the outline an equal amount on all sides.

The pads were extended slightly, to suit the soldering iron method, since the iron tip needs to touch the pads.

image

Speaking of soldering iron tips, for this scenario, a pointy tip works great, because it needs to get right into the corner between the component body and the pad. I used the tip shown in the drawing below, which to all intents and purposes is a pointy tip, since the dimensions of the end are so small (0.4 x 0.2 mm).

Any decent manufacturer soldering iron (JBC, Weller, etc) would work great, provided small tips are available. An integrated heater in the tip can make a good difference too. A low-cost soldering iron that uses tips with integrated heaters is the Secure S60; it supports such small dimension tips, and also works with larger tips for more normal soldering too. Note that if you're using a USB soldering iron, then it needs to be grounded when soldering semiconductors.

image

As mentioned, I splurged some flux paste onto the pads. Apologies for the photo quality, it was done with a shaky mobile phone. There's not much to see anyway, it's just flux. At these dimensions, the flux paste will be a single blob, it won't be precisely dispensed just on the pads.

image

The flux paste was called MG Chemicals 8341 others would be fine too, I'm just used to using that stuff so I stuck with what works.

Next, the part was oriented correctly and placed inside the silkscreen box shape with the tweezers, flat against the board. With the other hand, the soldering iron (with a small amount of solder on the tip was dabbed into the corner to touch the pads and the side of the component. Surface tension allowed the metal parts to retain just enough solder for a nice fillet, although really you're flying blind at this point, because the flux covers your view. Anyway, you can trust science at this point, and just dab in the general corner areas. 

Once that's done, the board is cleaned. I used isopropyl alcohol and a brush (it is very expensive for a brush, but it is brilliant, and will last for a decade or longer). The cheap brushes on Amazon are hopeless compared to that one, and then I used a wipe to mop things up.

Finally, the fillets of solder need to be inspected. You could use a mobile phone camera if nothing else is handy. I used a normal camera here:

image

As you can see, it's a bit wonky, but nevertheless all the pads are soldered with no shorts, and that's all that matters for this part. 

image

Summary

With not a lot of effort or cost, it is feasible to solder quite tiny parts, relying on surface tension and flux to manage the distribution of solder with no short circuits. Still, a fairly decent soldering iron and tip is needed, but they are not necessarily expensive these days. Particularly for a temperature sensor, you might want low thermal mass, and so using a small part might make the difference between merely an adequate solution or a (hopefully) better performing one. 

Thanks for reading!

  • Just realized the STS40 temperature sensor has the same footprint, and same I2C commands as SHT40 (combined temperature and humidity sensor), which is useful in case one or the other is out of stock!

    The code for it is really easy. To start a measurement, just write a single byte 0xFD to the sensor:

    // start temperature measurement. Wait 10 msec before reading the result.
    void start_temp_meas(void) {
        uint8_t buf;
        buf = 0xFD; // hi-res measurement command
        i2c_write_blocking(i2c_port, TEMP_SENSOR_I2C_ADDR, &buf, 1, false);
    }

    To read the temperature:

    // read temperature measurement
    float read_temp_degc(void) {
        uint8_t buf[3];
        uint16_t raw;
        float degc;
        i2c_read_blocking(i2c_port, TEMP_SENSOR_I2C_ADDR, buf, 3, false);
        raw = buf[0] << 8 | buf[1];
        degc = ((float)raw) / 65535.0;
        degc = degc * 175.0 - 45.0;
        return degc;
    }

    I've run some tests, and the measurements are super-granular. Each of these measurements were taken half a second apart.

    image

    Although all this is fine for board temperature measurements, measuring ambient temperature is a different kettle of fish. It really needs the sensor to be very far from any circuitry even if it's slightly warm, and there are the following guidelines in a PDF (screenshot below). My measurements are higher than room temperature by almost exactly 1.0 degree C, if the board is kept a couple of inches away from the microcontroller,but I think even conduction through pin headers is making an impact. If the board is closer to the microcontroller board (pi pico), then the measurements are 2.5 degrees C higher. Amazing what a difference that distance makes, considering the Pico isn't consuming a lot of power. This is without any enclosure so far.

    image

    I've decided to try to 'decouple from heat sources' (point 4 in the guidelines above) by making the traces even thinner and to narrow the PCB bottleneck a bit further. 

    I've tried to address points (1) and (3) by clearing up more space for any enclosure to be shaped around it and to provide decent access.

    image

    The STS40 could be powered down between readings, if it were powered from a GPIO pin rather than from 3.3V. However, I think a separate I2C interface (e.g. software I2C implementation) might be needed as well, to set the I2C bus low during power-down, which is no good on a single I2C bus if other peripherals need to be accessed. I'll consider that if the improvements described so far do not help as much as hoped.

    It is definitely a bit of a challenge getting good ambient temperature measurements. Anyway, at least now I know that with first board layout and with no enclosure, the error is approximately 1 degree C provided the sensor is kept a couple of inches distant away from the Pico board.

  • in reply to beacon_dave

    Nice idea! Such tools are really useful. I have a pre-heater, not quite the same as a hot plate since it doesn't reach reflow temperature (at least not without a lot of effort).

    The hot plate looks brilliant and very cost-effective way to solder these sorts of small parts.

  • The thickness of the PCB in those close-up photos really illustrates the scale you are working with. 

    Interesting to see the different techniques as components keep getting smaller.

    I was interested in Katie's mini hot plate technique in one of her element14 presents videos.

    image

    Looks like it could be a useful technique for some projects. 

  • in reply to genebren

    Hi Gene, 

    Thanks!

    I think there's an element of that occurring, but for hand-soldering I figured that was better than not having any outline, since it's very hard to align and hold in place a 1.5 mm part without any surrounding visual reference. Good point that it could be slightly larger. In the past the silkscreen had enough inaccuracy that even if it's slightly larger, since it's still sub-millimeter, one edge may touch (unless it's spaced out a lot, which is common with machine-placed parts). But from visual inspection, these boards seem to have an excellent silkscreen

  • Nicely done!

    I noticed that the silk screening on the PCB might be slightly lifting the component.  I find that on some tiny parts, I either leave off the component outline or enlarge it sufficiently keep it from allowing the part to drop through and make good contact to the PCB.