You are reading content from Scuttlebutt

Fired up the #offgrid freezer for the first time! At 200 PV watts incoming, it's running happily, no voltage sag to speak of. And it runs pretty quiet, though I will still have to get used to barely perceptable mechanical noise in my living space.




I overrode my automation to run it despite power not being great, and when I looked up from coding, it was already well below freezing! Much faster cooldown than expected, granted it's empty. Gonna have to enable computer control and let it heat up before I can begin using it now.


Old fridge, meet new fridge.


The old fridge was packed by that, new one is empty even with 7 gallons of added thermal mass.


Course I could put much of this in the fridge..


Tho it's all pretty safe at room temp. I did put the maple syrup in though.


@󠁪󠁯󠁥󠁹Joey Hess It would be interesting to understand the what and where of thermal leakage in these sort of units, and whether that could be easily reduced.
We have grid power, so it's not really an issue in the general scale of things, but for offgrid, I would like to know.


@󠁪󠁯󠁥󠁹Joey Hess Also, FWIW, the pipfruit industry here uses water dumps to float fruit out of orchard bins.
The obvious reason for this is that it is far gentler than tipping.
But the other big benefit is that it results in an initial and very efficient cooling of the fruit, so reducing the refrigeration load on the coolstores.

There may be some beneficial application of this idea in offgrid refrigeration?


Temp graph for today. Can see where my software switched to a more aggressive run mode as the rain cleared up later in the day. Small problem with dipping slightly below zero before the software turns it off, which will need to be dealt with.


The heating and cooling slopes are kind of sharp in the heat of the day, with it cycling twice per hour and running 15 minutes each time. And it was not super hot day.. 26 C.

But it was mostly empty still. Now I've upgraded to 17 gallons of thermal mass (cold spring water) plus a lot more veggies, and I hope that will level out the slopes more. In any case it looks like it'll get through the night without running and still stay well cooler than the propane fridge.


Last week of fridge data (sensor was out of the fridge on the 15th).


@IBob if you look at fridges designed for offgrid, like they have significantly thicker walls and lid with more insulation and thermal mass.

They're also 3 times the price. The one I linked to uses 140 watt-hours per day. It will be interesting to see how many watt-hours per day my solution uses, initial indications are 300-500. I suspect it's cheaper to go my route given current solar panel prices.

What I really want to avoid, which AFAIK commercial offgrid fridges are not designed to avoid, is running from battery power overnight.


@󠁪󠁯󠁥󠁹Joey Hess
Since you have an excellent logging system, It might also be interesting to temporarily wrap the refrigerator with blankets or some such to see what difference that might make. Anything to minimise or reduce air contact and movement with the outer shell??

User has chosen not to be hosted publicly

what are you planning to freeze in the long run? CSA meat or something?


@emile air temp currently, I was thinking the same about monitoring the ballast temp tho, and my probe is luckily completely waterproof..

It's tricky though; getting the air to freezing even if the ballast is not there yet will trigger lettuce leaves to crystalize. Luckily I have a lettuce bed, so I can insert control leaves while tuning it. ;)

@maggie refrigerate my food.


What I really want to avoid, which AFAIK commercial offgrid fridges are not designed to avoid, is running from battery power overnight.

Couldn't you just add some thermal buffers in the form of high heat capacity material? E.g. some jugs of water?


@juul yeah, I have 17 gallons water in there, could add a little more, but it starts getting tight for food. Found some stacking plastic bins that fit perfectly to hold it since more than a few gallon jugs gets messy.


Still working on organization too; the corner tower-of-cheeses is particularly handy.

It kept cooler overnight than my propane fridge used to, despite the ballast having just gone in at 8 pm and only being spring-cold, not fridge cold.



Moved the sensor into the water and am getting much less steep slope on cooling and much longer cooldown runs.


The water was 10 degrees warmer than the air. So far the air temperature is staying above freezing, although the walls do get a little frost on them since the coolant circulates there. It would be better to put the thermal mass around the walls.


15 degC seems a bit high for something that is meant to be a fridge. You may want to look into some phase change materials that work above water's freezing point. There are eutectic solutions that will unfreeze at around 2-5 degC, thus buffering the temperature at that point as they absorb heat during the phase change.

Here is an example of a commercial product, although it looks like they suggest taking them to 5 degC below their Eutectic point.

Perhaps try using an oil that solidifies in the fridge? I haven't been able to find any DIY solutions online but I'll keep looking.


I would suggest it's going to be a little difficult to tell what's going on during this pulldown phase: presumably that product went in at elevated temperatures, and there's a lot of it?

For blastfreezing of packed and cartoned meat, rapid air movement round individual cartons is an essential part of the process, and what we see is is an initial period which shows very little temperature drop (as the heat migrates out of the product) then a sharp drop once it is frozen. This is typically carried out with -20 to -30'C air and takes 1 to 2 days.

Stacking the same product in a -25'C coolstore (minimal air movement) will take weeks to achieve the same result.

I like @nanomonkey 's eutechtic solution in this setup, if a suitable medium can be found.


Fun fact (and very temporary thread drift, I promise!):
The first shipment of frozen sheep left NZ for the UK aboard the sailing ship Dunedin in 1882:
Makes ya think..........


Peanut oil has a melting point of 3 degC and is fortunately cheap to buy in bulk.

Your other option is to actually use the freezer as a freezer, and use propylene glycol pumped from the freezer into your old fridge to maintain the desired 4 degC fridge temperatures. In this case you would want to freeze a large mass in the freezer during the day and use it as a cold battery for your fridge.



Veeery interesting. Ran for 5 hours straight, at which point the water had cooled down from 13.5C to 9C and the air was holding fairly steady around 0C. Then with the fridge off, the water has continued to cool down for the past 5 hours! So the walls must have gotten quite well chilled down (probably well below zero) and are slowly cooling the water. Meanwhile, the air has slowly warmed to 5C or so.

I wonder what will happen over the next 8 hours before it turns back on...


@󠁪󠁯󠁥󠁹Joey Hess who would have thought monitoring an offgrid freezer fridge on another continent would be so enjoyable!

Just where on/in your water is your temp probe?

First, liquid water (obviously) is subject to convection.
Second, it has relatively poor thermal conductivity: there was an old lab demo where you sunk ice (with coarse solder wire) in a test tube of water, tilted the tube and boiled the water at the top of the tube over a bunsen burner, while observing the ice remaining ice (or only v slowly melting) at the bottom.
And I have also seen a long shallow steam bath with multiple electric elements boiling fiercely, while the temp probe stupidly mounted just below them never got over about 70'C.

May be a contributory factor here?


@IBob probe is currently a few inches deep into ~1 foot of water, a few inches from one of the walls.


@󠁪󠁯󠁥󠁹Joey Hess if you weren't off grid, I'd suggest using a 4" diameter PVC pipe connected to a USB fan to move off of the bottom of the fridge up to the top. I do that in my fermentation chamber when brewing to help minimize temperature gradients.


@Mark good idea, though unfortunately this introduces a heat source.........(
@󠁪󠁯󠁥󠁹Joey Hess I just enjoy trying to visualise system behaviour from historical data.
I've done a lot of industrial refrigeration, blasts, precoolers, stores, tunnels etc, but all of those are with our probes deliberately positioned in moving air (with the exception of multiple wall probes in large coolstores, where we are wanting to know if we have any 'warm' sides/end/corners.
As @emile said, probes in still air are fickle and very dependent on location: if your probe is close to or over something that is behaving as a relative heat source, then you get an elevated reading: in food processing rooms with temp compliance issues, we sometimes have to set up probes positioned away from heat sources to avoid this local effect.

In the case of your fridge, I am mostly interested to see how it behaves once the contents are pulled down to temp.

Thanks for the ongoing insight/entertainment!


@Joey Hess I'm not sure what your target temperature is but if it's not 0 C but is between -46 C and 17 C then you can use a cheap trick to improve the performance of your water jug buffer. By mixing glycerol and water in different ratios you can adjust the melting temperature to anywhere between -46 and 17 C. If you adjust the jugs to melt just above your target fridge/freezer temperature then the temperature will stop rising when they hit that temperature and stay stable until the water-glycerol mix has all melted.

There is a bit of a trade-off involved:

  • Enthalpy of fusion Glycerol: 200.62 J/g
  • Enthalpy of fusion Water: 333.55 J/g
  • Specific heat capacity of Glycerol 2430 J/g/C
  • Specific heat capacity of Water 4148 J/g/C

So you might want to take this into account when picking a target temperature.

You can find the melting point for different ratios of glycerol to water in a table in this paper from 1925 (Freezing Points of Glycerol and Its Aqueous Solutions by Leonard B. Lane).


Oh oops. Correction:

  • Specific heat capacity of Glycerol 2.430 J/g/C
  • Specific heat capacity of Water 4.148 J/g/C

Quite the difference :stuck_out_tongue:



Air temp in the fridge is approx 8C this morning. It took 8 hours after power off for the different parts to reach thermal equilibrium, and now the water is doing its job keeping it from heating up too quickly.

This seems pretty good, and if it can start running again before the water gets above 10C or so, it should be able to cool down further today.


@juul this project is kind of like a NASA mission, in that it's a success once it clocks a few hours of science on the surface of mars / temperature consistently lower than the propane fridge. Beyond that, 0-10C is a fairly arbitrary goal I coded into the software, if it continues in that range, great. Of course it would need to be consistently 0-5C for this to be generally acceptable by modern fridge standards.

Glycerol seems to run around $10/gallon in bulk, may give it a try once I get a better idea of the target temp.


Glycerol sadly has the tendency to supercool,so once all of your solution melts you will lose all of your seed crystals and then you have to drop down to a really low temperature before it will get back to a solid.

@juul, note that glycerine is denser then water. A solution of 95% glycerine and 5% water at around 5 degC is about 1.25 g/mL, so by volume it would closer to 3/4 of the capacity. Source.

Of course water won't go through a phase change above 0 degC, so all told the glycerine might prove to be a better temperature buffer if you can get it to actually freeze. Also, glycerine is a poor conductor, so if placed along the walls where the cooling elements are located you could reduce the temperature fluctuations that occur when the unit turns on.


@nanomonkey how low to re-solidify glycerol?

Looking like I should have no problem getting everything down to 1C most days as long as I space the food away from the cooling walls.


@IBob on temp probe placement, I think I know why it was seeing such sharp swings when outside the water. I had the wire coming down along the inner fridge wall corner, touching it for a foot, and then the probe was pulled out an inch from the corner. So, it was both very close to the walls, and wire's contact would suck more heat away when the wall cooled down.

The spring thermometer I have mounted in there to measure the air temp now avoids this problem, and has been staying just above freezing during long compressor runs despite frost buildup on the walls.

So, I wonder how I can mount the temp probe such that it gets the real air temp without this problem..?

It could come in along the top of the hanging basket, and so not contact the wall, but I want the basket to be movable. I suppose I could make a mounting rail for it.




@juul, note that glycerine is denser then water. A solution of 95% glycerine and 5% water at around 5 degC is about 1.25 g/mL, so by volume it would closer to 3/4 of the capacity.

I hadn't thought about that. Good point.

Of course water won't go through a phase change above 0 degC, so all told the glycerine might prove to be a better temperature buffer if you can get it to actually freeze.

It definitely still will be. The amount of energy absorbed during thawing is still around 150 J/ml vs. the ~4.2 J/ml needed to raise the temperature of water a single degree.


@Joey Hess Cool. Yeah I don't think you'll be able to get it cheaper than that but you might know people who bought a soap-making kit, used it once and filed it away in the basement. I see those old soap-making kits at thrift stores fairly often.


@󠁪󠁯󠁥󠁹Joey Hess my experience has been that selecting appropriate transducers, siting them correctly and mounting them well is half the battle with automation.

In refrigeration, stable systems are designed by appropriately sizing the components, and running them at pressures and with valving to achieve the required temperatures. Lotsa folk don't understand this: they tend to think that if you want colder, you just put in a bigger compressor etc.

You have an interesting challenge here, as the refrigeration in your unit is specifically designed to freeze: the evaporator part of the setup is pressure fed and jetted specifically to yield temperatures below 0'C
What that means is that the controls must intervene more agressively than in a refrigerator, which will be optimised to run over 0'C
(Okay, that's theory, certainly applies with industrial refrig, where we sometimes have to 'de-tune' repurposed cold spaces. I don't know to what degree it applies to domestic setups, but it will be a factor.)

One way to look at this might be as follows:
The air in the box has a minute thermal mass when compared with the steelwork.
So we can expect the air to follow the temp of the steelwork, provided the two are in contact.
So maybe identify what parts of the wall and floor are cooled by the refrig system (perhaps the upper walls, but I don't know), bond a temp sensor to the steelwork there, and control off that.


Moved the sensor so it's attached to the underside of the lid and dangling down an inch.

Might get better values this way, although now if I'm stood staring into the fridge, the sensor will read the outside air temp. And of course the lid is the warmest part by far so it's reading too high now.

Getting the feeling I'm going to want at least 2 sensors eventually..


PS More probes is better:
Old coolstores had maybe 2 probes. Nowadays we put in more: walls, by doorways, top and bottom of evap. And every time we do that, we go 'oh look...we didn't know it was doing that.'

PPS Since you are 'detuning' your refrig to stay above freezing, there are phases where you may have to introduce some timers to prevent the refrig short cycling, which is both inefficient and hard on the gear.


PPPS for repurposed refrigerated spaces where the refrig is too big/cold/aggressive, and assuming we can't get it by adjusting valves etc:

I'm going to have to deliver the refrig in a series of pulses to minimise temp overshoot (actually undershoot).
I will need to approximate the pulse size (due to delay in response time).
I'm trying not to short-cycle the gear.

So, I will deliver a pulse as required, and the duration of the pulse is some reciprocal of the temp rise time: fast rise, bigger pulse.
I will impose a minimum pulse value (since a pulse below a certain value provides no refrigeration)
and probably a max value to keep things toned down through door openings etc.
And I will impose a minimum pulse to pulse time to avoid short cycling.

An alternative to this is to have the pulses at set intervals, with the pulse width proportional to the the temperature error (actual temp (PV) minus setpoint (SV)).


@IBob no worries about fast cycling, I have it rate limited to 3 minutes.

BTW, I've read about chest freezer conversion kits that are so badly made that they cycle so fast sometimes that the relay humms..


From the same source

In spite of the tendency of glycerol to supercool, it can be crystalized by the use of seed crystals, or by cooling to -50 C or below and then slowly warming to about 0° C. Approximately one day is required for crystallization to be completed. The crystals are orthorhombic and very deliquescent.

I'm not sure if the seed crystals need to be glycerine, or what.

Did you see my note about Peanut oil having a melting point of 3 C, I found another source that says -2 I think it might vary. I found several sources that say Grape Seed oil melts at 10 C, and I know that I've purchased some olive oils that were solid in my fridge, although some just get cloudy.

The beauty of oil is you can always cook with it.


^^ @󠁪󠁯󠁥󠁹Joey Hess ^^


@nanomonkey Hehe, natural oils would probably have to be fractionated to get something with a specific melting point but that's a cool idea and I guess you could just experiment until you found an oil that worked.

It makes sense that glycerol doesn't want to crystallize but I was under the impression that this was only the case for very pure glycerol and not glycerol/water mixes? Now I'll have to do some experiments. Luckily I have glycerol and a freezer at home.


@nanomonkey @juul ok, I put a gallon of olive oil in a cold corner of the fridge, let's see if it freezes.

Course, oil has a specific heat much lower than water, I dunno how much energy it takes to melt it.


Moved the sensor again, the lid temp was too warm to be useful. Now it's coming across the hanging basket, right next to the dial thermometer. Which I now know reads 5 degrees C or so cooler than the sensor in the same location. (But the sensor wire might be conducting warmth from outside or something, so I don't know for sure which reading is more accurate.. Will add a second analog thermometer to calibrate..)

Interestingly, you can now see every time I open the fridge door, because the temp rises half a degree or so, and then slowly drops back down, but never quite to as low as before. Each door opening costs maybe 0.2 degrees in banked cold.


(I was just comparing sensor readings, not snacking all evening, honestly! :smiley:)

Otherwise behavior last night was more or less as expected, it was still cooling off the air until 3 am and then crept up slowly.

User has chosen not to be hosted publicly

@kas I've been measuring different ways at different times trying to find something that works well. A small quantity of water around the sensor may be my next attempt.


Liking the shape of this new curve with the sensor in the basket. Not to sharp and looks to be leveling off around zero.


Also, I have a fancy new line on the graph for hours run per day! That was some hairy FRP code to write to accumulate that, especially because the value has to survive restarts of my FRP program.

(Temp sensor has read consistently 5 degrees higher than the dial thermometer all along this curve.)

User has chosen not to be hosted publicly

Have done lots of simple probe calibration checks. It's all in how you make the ice slurry, which is simple:

Take enough (party) ice to twice fill thermos (or low thermal mass container).
Fold ice into rag or cloth to contain it, place on firm flat surface (floor) and beat it with mallet, hammer held sideways or some such. The aim is to smash it up as small as possible.
Fill thermos/container with smashed ice, including the fine 'snow', tamping it down.
Now add just enough water to make it look watery. The slurry should still be more stiff and solid than liquid at this stage. If you put in too much water, pour some out and pack in more ice.
Leave for 5minutes.

The water in the slurry will be 0'C (or close enough for Government work!)

We use this, along with a calibrated probe, for checking probes in situ to 1 decimal place.
It is very accurate and readily repeatable, provided the ice is smashed small, tamped down, and minimal water is added.
The calibrated probe tells us when the slurry is going off (rising in temp, which happens after approx 30minutes and multiple 6mm SS Pt100 probes), it can then be renewed by pouring off some water, adding more ice, stirring and leaving for 5min.


I should have be clear...that we don't pour off the 0'C water, we immerse the probe in the slurry.

In pipfruit stores, if the fruit freezes, the cellular structure is damaged, so the produce is held just above freezing. Some stores (notably large controlled atmosphere sites) have literally hundreds of temp probes for control and monitoring. Zero point calibration of these is checked annually using the above method.
A 6mm SS probe takes approx 2 minutes to stabilise once immersed.


The olive oil is a no go, didn't freeze despite the temp getting very close to 0. Will try some other oils later.

Found a crate that fits well in the center and moved the produce in there, so it's kept at least an inch away from the walls. Hoping this will let me run it more aggressively down to 0.

Intermittent rain and clouds today, but it stayed mostly around 1-3C and still managed to run for 4 hours.

User has chosen not to be hosted publicly

Last night it was 1 degree colder in the fridge than the night before, and outside temp was a tad warmer. So I think it's still cooling down a little bit as a whole.

All food is happy except for a zuchinni that got buried in a corner and frostbitten by the walls.

I've started looking more at derivatives than instant numbers, and at longer-term data. Here's daily power production before and after the fridge was installed. You can see it's gone up the last 3 days with the fridge's load. Before that, PV production would drop off to a low rate once the batteries were full, often by 11:30 am. Now, the fridge keeps banking solar power.




Temps over the past 2 days. I'm now pretty happy with its low temperature cutoff point; mercury thermometer in the veggie compartment consistently reads just above freezing after cutoff.

Overnight temps are well within the range I'm comfortable with, though I do hope to push them down into more socially-acceptable ranges. Also, once I stop opening it in the evening to check readings and reorganize it and stuff, it'll probably hold cold better..


@󠁪󠁯󠁥󠁹Joey Hess looks to be still on pull down?
I was trying to think why I like this stuff so much. I think it's to do with the pleasure I get from understanding and cooperating with my own environment to make things work well.

As distinct from standing in front of a glass door expecting it to open...


Gottit: the word is agency.........)


Its temp this morning was 7.3C, which is more than a full degree colder than yesterday morning. Despite the weather being a little warmer again. @IBob is right, it's still in cooldown.

A thermometer 2/3 of the way down (midpoint on the stack of water containers) was at the same time 3C!

So, the water is keeping a very good amount of cool overnight, the heating I'm seeing is mostly of the top layer of air. And I have a compartment already that is probably staying consistently below 4C (though also probably sometimes below 0C), in the "well" where the water tanks end. Currently I'm mostly only keeping some cold drinks and milk down there, but anything I want in that range (meat?) could go there.

My plans for the next week or so are mostly to observe it and do minor improvements to the control software. Longer term, I have a bunch of ideas to try..

  • Reorganize so the food is stored below the level of some of the water, so that it sits in a pocket of cool air under a ceiling of warmer air. This probably involves water tanks up the side walls with the food in the middle. Also insulating from minus-degree walls while cooling. The disadvantage is it will be harder to access the food, while right now it's mostly a reach-in operation.
  • Or add a small circulation fan to mix the air, instead of reorganizing the water tanks.
  • Insulate the lid some more, but I doubt this will have a big effect.


  • On a partly sunny day, there is still solar capacity going to waste in the half of the time when the fridge is too cold to run and batteries are charged. My old propane fridge can run on 12v DC or AC, so hook it up with a relay to turn on at those times, and try freeze water in it. If that works, I'd have a freezer compartment; I could also try to circulate the cold over to the fridge at night.
  • Try switching it to freezer mode, so it can run all day long, and pipe the cold over to a free junk fridge or my old propane fridge.

I forgot:

  • Accurately measure a conventional fridge set to a conventional cold level, and find out what the temps really are in the different areas.

Which if anyone would like to do...


So I've been thinking about this project far to much lately. I'd like to say it was mostly while drinking coffee sitting on the porch petting the dogs, but there was a decent amount of time spent online researching Phase Change Materials and insulation. I found a company that produces PCM in a variety of target temperatures who is willing to send samples, but they include a surcharge of $250 for the process of making anything under a ton. Hrmmm.

@󠁪󠁯󠁥󠁹Joey Hess, have you ran the freezer overnight as a fridge and looked at what the battery usage would be? If the unit only uses 600 watts a day as a freezer, it may only take say 150 watts of battery power to maintain temps at fridge levels at night.

As you've mentioned, using it as a freezer might be most effective as a "thermal battery" if you actually freeze some water and use the frozen material to cool an insulated cabinet or your other fridge. You could use insulated hvac ducting or insulated copper pipe to transfer cold air or a transfer fluid between the two. During the winter when there isn't much sun and it's snowing you could use your cold outside air as a cold source.


@nanomonkey I tried running it on battery power alone at one point; it sagged my batteries almost to the LVD. That sag persisted after the compressor surge IIRC -- only ran it for 5 minutes probably. I'd need more batteries, probably 2x-4x. This makes sense since my current battery bank is pretty well tuned for a 40 watt night load.


On the subject of batteries, I've actually had lots more battery power at night after installing the fridge. They get a better charge with the extra power flowing through the system during the day.

I had observed my charge controller undercharging batteries when load was minimal before, so I'd come home from a trip and that night would have only half the usual power. When that happened the absorb stage spike was very tall and very short, and the batteries sat in float all day at a relatively low voltage.

I've read that higher lead acid float voltage can improve their charge, but of course it also risks damaging them. Hopefully whatever effects the fridge is having indirectly on the batteries are good for them long-term.


@󠁪󠁯󠁥󠁹Joey Hess This all looks very promising.
For those of us following this wrestle with realities.......could you please summarise the current physical load (produce and water ballast) distribution and probe/s positions? Thanks.

And should you decide to add a fan, it may be worth presenting both cold and warm air to the fan (if fan is at top, conduit some cold air from the bottom, or vice versa) so that the fan is vigorously chopping and mixing the two temps.
The reason for this is that fluids of different temp tend not to mix as readily as we imagine they will:
on one job I had plumes of warm and cold air travelling down ducting, round corners etc yet remaining quite separate. This was for a big proving room, we had 6 such ducts with probes mounted through the walls giving different readings depending on position in airflow. So we introduced various baffles and spirals to tumble the flow, but with only partial improvement. The source of the problem was heating elements in the duct that did not reach fully across it (allowing air at one side to pass unheated) and we finally fixed it by narrowing the ducts so that all the air had to pass through the heating (fix the problem, not the symptom!)
In that instance, the fan was before the elements: I doubt we would have had the problem if the fan was after the elements.





(Sad state of asparagus is because it spent too long last week in the old fridge.)


@󠁪󠁯󠁥󠁹Joey Hess thanks. I hadn't realised your water ballast was so big, and all in one lump (not that there's anything 'wrong' with that, just that I'd visualised it differently).
What is the empty volume of the machine?


8.7 cubic feet

(I may have drawn it a little taller than it is.)


Hi Joey, I had a MPPT in my cabin for a couple of years in our sailboat, while we lived onboard that was rather annoying. I remember turning the solar panels of a couple of times because I wanted a late morning day. I have a failed freezer installation that needs a redesign on the boat. I'll love to pick your brains with it in the future. I'm not having much time to look after her now.


Volumes continue to be deceptive for me.
The freezer volume looks bigger than 8.7 cubic ft...
while I had to check twice that 17gal of water is just 2.27 cubic ft.......


@󠁪󠁯󠁥󠁹Joey Hess
Does the floor of the unit have a raised portion at one end, with the refrig gear underneath? In a sense the ballast/produce positions are wrong way round (though probably easier to control that way). I'm wondering what the dynamic would be if the ballast were higher up (but at one end, to give reasonable access to the produce.

Thank you for letting us watch, and comment on, your freezer/fridge project.


May have just fried the compressor. Had just started a manual run at 6pm, with 64 watts coming in and battery voltage 25.9, when I heard something unusual (dunno what; radio was on) from the fridge and the compressor shut off. Its run light was still on, tried one more brief power cycle and still not turning on.

I suppose I'll know tomorrow morning when there's full solar power, if it doesn't come on.

Smelling the compressor ventilation area, it does not smell promising.. :worried:


Schematic has a motor protector, and the run light would still be on when that is open, so I'm hoping it's a fuse..

User has chosen not to be hosted publicly

Bit difficult to remember, but based on logs and iirc, it was already running when it stopped, probably not a compressor start problem.

An AC voltage sag from the inverter blowing a fuse seems more likely.. I have not checked how stable its AC signal is, although it's true sine wave.

User has chosen not to be hosted publicly


Motor housing is warm to the touch, but not uncomfortably hot.



No obvious fuse or damage.. Also no lingering electronics fry smell, it could have just been new electronics offgass I smelled. Hmm.


@󠁪󠁯󠁥󠁹Joey Hess with big compressors, we normally limit starts to 6 or 8 per hour: I do this by putting in a min start-to-start time.
While I'm at it I put in a min stop-to-start time to avoid short cycling (also to avoid really short cycling if I stuff the code up).
A freezer running under normal continuous on-grid conditions is self- limited by it's own normal thermal.cycle. However, in your case, it may be good to impose something like this to allow stuff to cool down?

User has chosen not to be hosted publicly

FWIW here's how long each run cycle lasted today. (Units are hours.)


The failed start was after the last run shown there so it had just finished a run of approximately 8 minutes when I forced it back on.


And yeah, preventing too short runs and too close runs is why I have been tracking that particular data... It has a hardcoded 5 minute min time between changes to power state currently. But I manually override that for the failed run.

User has chosen not to be hosted publicly

Well, if it doesn't start, we'll get a great test of how long the ballast holds its cool...

I'm more worried about the salmon. :-P


Whew, verified it was thermal shutdown, it runs again now.

So, I need to

  1. prevent my override from turning it on immediately while still letting me overridde power to off immediately
  2. probably increase the 5 minute cooldown period to 15 or so
  3. get a temp sensor on the compressor so I can more smartly prevent running it too hot (have a bunch more temp sensors already on order)

Salmon? We'll be right over.....................)

Thinking further about short-cycling (which cannot normally occur with conventional on-grid running): it may be that the compressor then has to start against residual refrigerant pressure (which in conventional running would have bled off). This would make for a harder start.

The whole story with bigger compressors (which may be totally irrelevant here) is that we limit starts per hr, as noted above;
We do not start them loaded (that is, they are not starting against any great pressure);
And in the case of one brand (Bitzer) we are told to run them for a given time (7min, I think?) once started to dissipate heat (though personally I think this is BS, and due to some early design problems they had).


@emile I can't find the compressor's specs anywhere by model number or QR code (decoded to timestamp: "200907186014118011408127"). Sure would be good to know what temps and duty cycle it's rated for.


@󠁪󠁯󠁥󠁹Joey Hess sounds good:
allow time to dissipate heat, and for refrigerant pressure to bleed off.

As I recall, some of these little units do run very hot?


Got a temp sensor mounted on the motor (using some high-temp, "sensor-safe", electrically insulating silicon paste from the auto parts store), and while I was at it built and installed a breakout board in an enclosure under the fridge, so I can easily plug in additional sensors.


Motor seems to run around 40C in normal operation, after it's been powered on for 20 minutes or so.


I'm pretty sure you'll be okay provided the cycles are adequately spaced, and the voltage is correct; these units are very durable. However, since the entire thing is sealed in a steel can, I wouldn't expect any overtemps to be apparent from outside until well after the event.

Basic temp safeties on industrial compressors are:
Oil temp
Discharge temp (the hot gases being pumped out).
Motor thermistors.

You could get the discharge temp by attaching a sensor, with plenty of thermal insulation round it, to the copper discharge pipe from the can. However, I doubt this would give you any useful information.

A couple of nice new low-stage Mycoms going for it:


Today's motor temps


Based on this, I've programmed it to not start when the motor is > 45C. Summer temps here are typically below 38C , and on rare occasions up to 44C. So 45C may not be enough of a gap on a blazing summer day, although the porch will be some cooler. We'll see. I could factor in the outside air temp, but want to keep this simple (hah!).

Also put in my own thermal shutdown at 70C which I'll adjust down when I have a better idea of the normal range, but I expect the compressor has a better one built into it so if mine triggers it's probably on fire.


@nanomonkey a quart of peanut oil froze..

It went in on the 23rd and loading it with groceries kept the temp higher than usual that night. But then last night the temp stayed under 6C and basically flat all night. Partly a cooler night, but I wonder if the oil is helping..



Wooohooo!! Would it be possible to put one of your temperature probs in the peanut oil, to see what temp its buffering at?

I've been in contact with PCM Products, who makes a variety of Phase Change Materials for buffering temperatures in homes, fridges or computers. The A4 material which buffers at 4 C and is made from "organic materials" costs about £4~6/kg when purchased in 1 ton quantities. I believe the price variation depends on how it's encapsulated, ie what type of ice packs you have them put the fluid in.

Of course peanut oil is about $37 for a 5 gal/35 lb this could be a much better DIY solution. And I'd much rather eat peanut oil on hand to eat then some random unknown fluid.


@nanomonkey what sort of figures do PCM they quote for the latent heat of fusion? It's unfortunate that it's so low for peanut oil.


@nanomonkey well, the fridge would need to get above its melt point. But, I'll try to determine the melt point experimentally at some point.


@󠁪󠁯󠁥󠁹Joey Hess @IBob

Here is the spec's sheet on the PCM materials. The A4 has a Latent heat capacity of 200 kJ/kg (water is 334 kJ/kg) and a specific heat of 2.18 kJ/(kg⋅K) (water is 4.19 kJ/(kg⋅K)). Generic vegetable oil having a specific heat range from 1.67 - 2 kJ/kg.

From this source, the latent heat capacity of peanut oil is 21.7 cal/g which is 90.8 kJ/(kg⋅K) which is half of that of water. Although, oils often melt over gradually over instead of at one eutectic point, so it's possible that a decent amount of thermal energy is stored in the range (note it's a per deg K change rate). This document gives examples of this phenomena in usage for cooling systems.


@nanomonkey this table seem to indicate that the latent heat of fusion of peanut oil is very low......assuming it is correct(???)


@nanomonkey ^^


@IBob It seems suspect that it would be so much different then olive oil.


@nanomonkey yep, does seem a bit suspect.
I'm not clear about oil anyway: it doesn't so much freeze as thicken, due to a variations in structure, or some such.


Added two more temp sensors just now! (Well, it took 3 hours, I had to modify the distribution board to add more terminals and stuff.)

One is a probe into the water tank. Currently reading -0.062 C. Which is very interesting, because that water is still liquid. Supercooled? Need to move the probe to not be on the bottom of the upper tank to get a more accurate reading, there could be ice between the tanks. I had some ice crystals in a jug of tea down in the cold well the other day, very refreshing.

The other probe is kind of a spare, I have it down near the bottom of the cold well right now. 0.187 C, and half an hour ago I noticed it read exactly 0.0 C.

Looking forward to some interesting graphs...


@󠁪󠁯󠁥󠁹Joey Hess What is the accuracy spec for these probes? Or what is the device, so I can look it up for myself?

It seems to me that in a sealed area like this, there is likely to be some top-to-bottom temp gradient.
I suspect it would be fairly minimal, except for periods after lid openings, or when a larger heat load is added. But it would be nice to know.

I have a thing about 'degree': in this case, the question in my head is not whether the freezer will be warmer at the top, but to what degree.


@IBob DS18B20

Comes factory calibrated to ±0.5C accuracy, but it is possible to get a better calibration; the readings are very stable and form smooth curves to within 0.05 or so.


According to accuracy varies quite a lot from individual sensor to sensor.



Result of the code change to not start the motor warmer than 45C, compared with yesterday before the change. Seems to have improved run durations well, although short runs are still possible in the morning before the motor warms up much, and I suppose it won't have the same effect in winter.


off for 18 hours


Very rainy day, could have run plenty, but the code prioritized charging batteries instead since it was not very warm, and I mostly agreed with it.

Good opportunity to see how the temps develop over time. I think having part at 2.5C after 18 hours off is very good, and it was all below 6C until noon. (Outside temp ranged 19-22C)

Now below 2C, after a mere 1 hour running today.


@󠁪󠁯󠁥󠁹Joey Hess So there is a 4 to 5deg split between the top and bottom of the fridge??? If so, that's huge!!!


The peanut oil is staying frozen up to nearly 8C, which is what the temp got up to overnight due to running so little yesterday.


So I took out the cold well probe and am trying this experiment.


Currently at 6.75 and has the consistency of KY jelly.


Let the peanut oil warm up until there was no ice left, then chilled back down to freezing.


Freezing point is somewhere between 7 and 4.5C, and my best guess is around 6.3, where both sides of the graph are flattest.

Certainly it seems to freeze higher than the 3C google finds.


Been grappling the past several days with a tricky situation for my software to handle. On a cloudy day, the charge controller gets the batteries to around 85% and then it starts reducing the watts produced. It seems to be in an absorb stage, although this CC only reports boost/float and claims it's in boost. Sometimes it will be in this state for hours.

Turning on the fridge at this point is desirable, because rather than producing 64 watts through clouds, it'll suddenly be producing 200 watts.


But, it could also be that the battery got to 85% full and then it got very dark out, and it's only able to eke out 64 watts max, and then turning on the fridge will start draining the battery at the worst time.

One way to solve this would be to install a light meter. Trying to find a good one that can be put safely in a good location is not easy. The best ones are i2c bus and so difficult to get far from the computer.
Could try to diy a 1-wire using

Or I could try to briefly switch on a load other than the fridge, to probe if more watts can be produced. Eg a light or heating element (perhaps the one in my old propane fridge).

I'm currently trying to solve it in software, by looking at all the data, especially the voltage, I hope to be able to detect the absorb state.

"Luckily" I have had plenty of opportunity to grapple with this due to a total lack of direct sunlight this week. The fridge is keeping between the lines, but only because I'm noticing this situation and manually overriding it to turn on.


@󠁪󠁯󠁥󠁹Joey Hess
First thought: this continues to be an interesting 'real world' study!
Second (tractor automation) thought: yep, I really don't like expert devices that do their own thing without providing full status info as to what they are doing and why.

My own inclination would be your second option, but mainly because I have a long term low maintenance attitude: a resistive test load of some sort could be simple, robust, and would last indefinitely.
On the other hand, a light meter, while probably requiring shelter from the elements, and periodic cleaning, would yield more data.

The perfect fix would be a CC that also delivered rich raw status info.............


Is your server too far away to use a Photoresistor? If so, it may be a good opportunity to use one with an esp8266.


Weird.. My temp sensor on the fridge compressor has begin to cut out when the motor is running. Works the rest of the time. Seems to cut out at higher temps, but not always the same temp.

Luckily everything works without that sensor..


Cutting out as in not responding to comms....or not delivering a realistic reading?


No comms at all.


A first look at longer-term data trends for the fridge...


At the start of this period, it was well chilled down, note that the water temp was just above 0C. Then a week with an unusual amount of rain and clouds, nearly worst-case situation. It was only able to run for 1-2 hours most days that week.

Looks like 2 hours of running per day is enough to keep the water temperature from rising, while with 1 hour of running, the water temperature rises by around 1.5C per day. (Given exterior temps in the mid-20's.)

The water is a good couple of days buffer to the temperature. I was satisfied with the overall temps in the fridge during that period; while in places they rose to the 8.5C shown in the graph, the place I keep meat and milk etc was below 5C except for a few hours. No problems with any food.

Once the weather improved, it took 2 days of long running (3 and 4 hours) to push the water temperature back down. As long as there's a good day now and then, the fridge seems like it will do well.


Yesterday was perfectly clear skies all day, and this is the graph of it.


You can see every time the fridge turned on, and how much potential energy is still being wasted on days like that. Bearing in mind that the actual curve of available power peaks in the afternoon and ought to go up to nearly 1500 watts..


Ran the old 3-way fridge from AC today to see how it performs. No problem running it at the same time as the chest fridge; it draws around the same amount of watts.


That is with the spare probe moved to a cup of water in it, until the end when I took it out to measure the final air temp.

Unsurprisingly, resistive cooling is a lot less efficient. Compare with the initial cooldown of the chest fridge at the top of this thread, which went to zero in half an hour. The entire loaded chest fridge cooled down faster today than the empty 3-way. On the other hand, it is silent and I can cycle it on and off with every passing cloud. Seems worth automating that and seeing if I can use it as a little freezer.



Join Scuttlebutt now