After hairballing an LM75A I²C temperature sensor to verify at least one of the eBay lot worked, a bag of SOIC-to-DIP space transformers arrived, so I soldered up another LM75A:
The SOIC chip pattern sits at right angles to the DIP pins, which took some getting used to.
The slightly defocused wire connecting pin 4 (on the IC) to pins 5, 6, and 7 (on the PCB) selects address 0x48.
So I flipped it over, soldered four wires (+5 V, GND, SDA, SCL) to the numbered pins on bottom of the board, made up a little header for the other end, wired a socket strip on the crystal tester board, plugged it in, and … nothing worked.
Turns out that the other side of the board carries a TSSOP pattern, which I’d neatly masked off with a snippet of Kapton tape, surrounded by eight numbered pins. Of course, those pin numbers correspond to the TSSOP pattern facing you, so they’re mirror-imaged for the SOIC pattern on the other side.
Soooo, the proper wiring for the SOIC pattern as seen from the TSSOP side has the pin numbers exactly bass-ackwards:
The insulation looked a lot better the first time I soldered the wires to the PCB. Honest.
Anyhow, when correctly wired, the LM75A worked as it should:
It’s snuggled chip-down against the top of the 125 MHz oscillator can, with a dab of heatsink compound improving their thermal bond and a yellow cable tie around the foam holding them together. The socket header is wired pin-for-pin to the DAC I²C socket directly above it.
The OLED temperature display shows 28.250 °C, because the oscillator just started up in a cool basement. It’ll eventually settle around 39-ish °C, where its output should be pretty close to the 125 MHz – 344 Hz value hardcoded into the source.
Oh, that’s a 3 mm amber LED next to the relay can: much less glaring than the white LED, no matter what it looks like here.
The mailing tube arrived with contents intact, although the USPS inlet scanning didn’t work and the tube pretty much teleported across several states without leaving any tracking data behind. The recipient suggested several modifications to the caps:
Review of user experience of tube end:
The ribs on the endcap are very good at holding the cap on, so much so that I had to use a prying implement to remove it, which cracked the flange.
Would consider less depth on the cap, and possibly another layer on the flange.
Some continuous process improvement (a.k.a OpenSCAD hackage) produced a swoopy threaded cap with thumb-and-finger grips:
The finger grips are what’s left after stepping a sphere out of the cap while rotating it around the middle:
That worked out surprisingly well, with the deep end providing enough of a vertical-ish surface to push against.
The two hex holes fit a pin wrench, because the grips twist only one way: outward. The wrench eliminates the need for a flange, as you can now adjust the cap insertion before slathering packing tape over the ends. Man, I loves me some good late binding action!
A three-start thread seemed like overkill, but was quick & easy. The “thread form” consists of square rods sunk into the cap perimeter, with one edge sticking out:
They’re 1.05 times longer than the cap perimeter facets to make their ends overlap, although they’re not tapered like the ones in the broom handle dingus, because it didn’t (seem to) make any difference to the model’s manifoldhood.
Not needing any endcaps right now, I built one for show-n-tell:
The OpenSCAD source code as a GitHub Gist:
One of the leather strap anchors on Mary’s giant haul-everything-to-a-concert(*) handbag pulled its rivet through the canvas fabric:
We knotted the strap around the zippered opening and completed the mission.
Of course, it wouldn’t have pulled through if they’d splurged on washers, but noooo too expensive:
Some rummaging produced a pan-head M3 screw of suitable length:
A slightly battered acorn nut was a special treat for the inside, with another washer to keep me happy:
That was easy!
Spotted this impressive array at an apartment building:
That’s just for one wing; the other end of the building has a similar installation. Each apartment has an electric stove and gas heat / AC.
Disabling the display by activating its powersave option reveals 60 Hz pulses from the USB port on the Arduino Nano:
Unplugging the USB cable, leaving just the +5 VDC power supply and coax cable to the oscilloscope, solves most of the problem:
A closer look shows some (relatively) low frequency noise remains in full effect:
Disabling the display while measuring the crystal seems sensible, although, to avoid surprises, a pushbutton should start the process. Unplugging the USB port puts a real crimp in the data collection, although that’s probably survivable with a USB isolator, one of which is on the way around the planet.
The remaining low-level chop requires more thought. Somewhat to my surprise, holding the Arduino Reset button down doesn’t change much of anything, so it’s not a firmware thing.
Those 10 µF coupling caps gotta go.
With the OLED dark and the USB carrying data:
Compare that to the first pass:
Tamping down the noise seems to reduce the overall amplitude variation, but it also makes the capacitor-in and capacitor-out curves more consistent. There may be other things going on that I haven’t accounted for.
The peak frequencies differ by 0.2 Hz, which is probably due to a few degrees of temperature difference. Obviously, it’s badly in need of a temperature calibration & correction.
A strip of NXP (née Philips plus Freescale, including the part of Motorola that didn’t become ON) LM75A I²C temperature sensors arrived from beyond the horizon. To see if they worked, I soldered thin wires directly to the SO-8 pins, entombed it in Kapton tape to prevent spitzensparken, and jammed it under the foam insulation atop the AD9850 DDS module:
This turns out to be easier than screwing around with thermistors, because the chip reports the temperature directly in Celcius with ⅛ °C resolution. Classic LM75 chips from National (now absorbed by TI) had ½ °C resolution, but the datasheet shows the bits have an easily extensible format:
Huh. Fixed-point data, split neatly on a byte boundary. Who’d’a thunk it?
There’s a standard Arduino library using, naturally enough, floating point numbers, but I already have big fixed point numbers lying around and, with the I²C hardware up & running from the X axis DAC and OLED display, this was straightforward:
Wire.requestFrom(LM75_ADDR,2); Temp.fx_32.high = Wire.read(); Temp.fx_32.low = (uint32_t)Wire.read() << 24; PrintFixedPtRounded(Buffer,Temp,3); u8x8.drawString(0,ln,"DDS C "); u8x8.drawString(16-strlen(Buffer),ln,Buffer); printf(",%s",Buffer); ln += 1;
The next-to-last line squirts the temperature through the serial port to make those nice plots.
Casually ignoring all I²C bus error conditions will eventually lead to heartache and confusion. In particular, the Basement Laboratory temperature must never fall below 0 °C, because I just plunk the two’s-complement temperature data into an unsigned fixed point number.
Which produces the next-to-bottom line:
Alas, the u8x8 font doesn’t include a degree symbol.
Given sufficient motivation, I can now calibrate the DDS output against the GPS-locked 10 MHz standard to get a (most likely) linear equation for the oscillator frequency offset as a function of temperature. The DDS module includes a comparator to square up its sine wave, so an XOR phase detector or something based on filtering the output of an analog switch might be feasible.
The crystal test fixture and amp huddle in front of the OLED display:
The 22 pF cap now sits across the relay’s NO contacts, so as to simplify measuring the total in-circuit capacitance. The LED turns on when the relay shorts out the capacitor, which has a 50% probability of making more sense.
The quartz tuning fork resonators have an ESR around 20 or 30 kΩ, so the off-resonance output should be down something like -60 dB = 20 log (24 / 24×10³) from the 150 mV input: 200 µV (-ish). It’s actually around 1 mV, suggesting plenty of blowby through the baling-wire connections hidden under that neat top surface. I think that’s why the whole setup shows only about 8 dB of dynamic range; more attention to detail may be in order, although the peaks probably won’t move all that much.
Anyhow, even though the AD8310 log amp module should be able to handle such a tiny signal, the MAX4255 amp provides 40 dB of gain (OK, just 39.8 dB) and rolls off the high end at 220 kHz as a side benefit of its 22 MHz GBW.
There’s way too much low frequency rumble at the amp output:
What look like grass is actually the 60 kHz resonator output: those big lumps & bumps are noise from this-and-that. The repetitive peaks and dents exactly 10 ms apart (the cursors span four of ’em) felt a lot like OLED refresh cycles and, indeed, went away when I yanked the display out. Pulling the USB connection eliminates another tremendous heap o’ noise, so there’s likely a ground loop (-ish) thing going on, too. This may call for a USB optical isolator, its commercial equivalent, or more eBay offerings. Getting rid of that junk may improve the dynamic range enough to keep me from doing anything drastic.
The AD8310 log amp input now has decent coupling caps, so it’s not seeing the VCC/2 bias, and I removed that kludged-in 50 Ω terminating resistor to present its full 1.1 kΩ input resistance to the op amp.