Small and crude GSHP for a Michigan garage

[RobertSmalls]

You should calibrate that against a known, such as a 0°C mixture of ice and water.

[pick1e]

Originally Posted by RobertSmalls

You should calibrate that against a known, such as a 0°C mixture of ice and water.

That's in the works, I just haven't encased them in waterproof probes yet. Supposedly they are accurate to ±0.6°C across the 0-100°C spectrum. But even with them calibrated at 20°C I can see the variance increase slightly at 4°C. I don't know if they are varying at different rates or if there is some hysteresis. That will be interesting to find.

I hope to make the probes by this weekend, then after that test the output of my heater for the borehole thermal response test.

How's the storm coming in NY?! We've got about a foot so far and 40mph winds. I JUST finished shoveling the driveway when my girlfriend's work finally called to announce closing.

P.S. With a ±0.6°C across the 0-100°C spectrum I should have no more than ±0.24°C between 0°C and 20°C, calibrated at one end. We'll see about that

P.P.S. I created my own "known" at room temperature for the first calibration by using the mean of the 6 sensors. I figured that was pretty fair. Not sure if I mentioned that. I calibrated each sensor by giving it an offset to move it in toward the mean.

[RobertSmalls]

Originally Posted by pick1e

How's the storm coming in NY?! We've got about a foot so far and 40mph winds. I JUST finished shoveling the driveway when my girlfriend's work finally called to announce closing.

Not bad. It was a quiet day at work, because most people had made up their minds the night before not to show up. I woke up this morning, hoping for a snow day. But all we got was 4" of extremely heavy, frozen snow + sleet mix. There was freezing rain on my way in to work, and snow on the way home. It's still snowing, so I should probably go shovel again before I turn in for the night.

[pick1e]

After administering a dopeslap I corrected my software to account for the increased resolution. This gave me better results but samples were still bouncing around. I played with the sampling technique at length and finally found the best way to get stable results. Basically I take 100 repetitive samples and average them- but the key was avoiding any unnecessary math in the processor until all samples were taken, basically:

allocate 6x100 array
for 1 - 100
  for 1 - 6 {
    get reading
  }
}

for 1 - 6 {
  average readings
  convert readings to human readable
}

I like these results a lot. On previous graphs the samples amongst the sensors diverged when the temperature diverged from the calibration temperature. Here they diverged when the temperature changed rapidly but converged again as the rate of change slowed. The variance between high and low samples was within 0.2°F at 70°F and returned to the same variance at around 0.4°F/min. The variance may be due to some hysteresis but could also be due to differences in convection in the crisper (I ran this test in the crisper to reduce uneven air currents). The maximum difference was about 1°F as shown by the black line, which still isn't bad in my opinion. However, I think I will try again with a heat spreader and see if the variance is reduced.

[RobertSmalls]

Looks so much better. Considering what you're planning to do with the sensors, I'd say the drop in effective sample rate is a good thing.

So, what do you think the temperature in your fridge is?

Hey, before you bury your six thermocouples in the yard, maybe you should log some temperatures around the house.

Suddenly, the Arduino Uno looks really cool, but its site is really lacking some important information. What's the sample rate and resolution of the Uno? If this $55 Arduino is half as good as National Instruments' $300 datalogger, I'll buy one.

[pick1e]

Originally Posted by RobertSmalls

Looks so much better. Considering what you're planning to do with the sensors, I'd say the drop in effective sample rate is a good thing.

There really isn't any change in the sampling rate. I'm manually setting a delay between samples just so I don't have more data than necessary. It takes the Arduino only a fraction of a second to average those 100 samples from all 6 sensors, so it's very fast.

Originally Posted by RobertSmalls

So, what do you think the temperature in your fridge is?

Well here's the portion of that graph after the sensors reach the temperature of the fridge:



So I can safely say the temperature averages about 40.25°F. Needs to be turned down a little yet. The sensors are guaranteed within 0.6°C from the factory and I gave each of them an error offset bringing them toward the average of the 6. The average of the offsets is 0.00 so I'm fairly certain of their accuracy.

Originally Posted by RobertSmalls

Hey, before you bury your six thermocouples in the yard, maybe you should log some temperatures around the house.

No problem. At $1.70 apiece I bought a whole fist full of them. Right now though I'm working on reducing the error from interference on long cable runs. At 10mV per degree I'm getting a lot of interference error.

Originally Posted by RobertSmalls

Suddenly, the Arduino Uno looks really cool, but its site is really lacking some important information. What's the sample rate and resolution of the Uno? If this $55 Arduino is half as good as National Instruments' $300 datalogger, I'll buy one.

As I mentioned before the sample rate is quite fast. I'll make a program to check how many samples I can take in a second but I'm sure it's way more than necessary. I have a delay time between samples programmed in, in this case I think it was every 5 seconds. It would take 30 seconds to change that delay in the program and upload it to update the Arduino say if you wanted to change the delay to 10 seconds or 3 seconds, whatever.

The analog to digital converter on the Arduino is 10 bits giving it 2^10 or 1024 steps. You have several choices for the reference voltage (analogReference) but for example the default is 5V. So it reads step 0 for 0V and step 1023 for 5V. In the case of a LM35 sensor at 10mV/°C, the output ranges between 0.02V and 1.0V corresponding to 2°C to 100°C. So that would only range from steps 4 to 205 on the Arduino. So 201 steps per 98°C is only a resolution of 0.49°C/step. To get the best resolution I used a voltage regulator putting out 1.2V as the reference voltage to get down to 0.12°C per step.

What I like about the Arduino is how universal it is. It can be repurposed to do whatever you want very easily. So I can use it now for logging data but eventually turn the same thing into a controller for multiple sensors, pumps, valves, whatever.

[RobertSmalls]

How's your hole?

The ground underneath all my snow doesn't appear to have frozen (worms still chilling out a few inches deep), though the floor of my garage is around 30°F. Have you observed frozen ground in your location?

[okpiddler]

I have used a 5k btu window ac to cool a bait tank by putting the evaporator coil in the water. It worked great even using the original T stat. I did not need it long. The fins (not on the fish) started corroding a little. I put it back as an airconditioner.

[pick1e]

Originally Posted by RobertSmalls

The ground underneath all my snow doesn't appear to have frozen (worms still chilling out a few inches deep), though the floor of my garage is around 30°F. Have you observed frozen ground in your location?

I hadn't even had a chance to check but I'm sure it was nice and toasty in my yard under a foot of snow even when it was around 0°F at night. Normally, though, we don't have the luxury of that much snow all winter.

Unfortunately my project is on hiatus since we are in the process of buying a home and moving very soon. Hopefully I will be back at it this summer. On the upside the new house has retrofoam insulation and a 90+ furnace. It has an attached garage but will still need to be heated to work in the winter comfortably. I hope to be back at it in earnest this summer.