airving

Recipes 

Ok - hot side liquid cooled, cold side will have air over a solid heat sink. That seemed to be a good plan to me as well (and still does - but again, not a professional here by any stretch). That unit looks pretty snazzy - I like the enclosed nature of it.

My plan had copper plate liquid cooling units (Swiftech Apogee GT CPU Cooler Blocks - originally intended for cpu liquid cooling) bonded to either side of the peltier. The hot side had glycol solution circulating to a radiator (Swiftech MCR320-QP Quiet Power radiator) with 3x120mm fans, and the cold side simply cirulcated glycol soln to a small tank for testing. I'm sure that proper (more than I had) insulation must be the key, but I wasn't able to maintain more than a 20 degree F drop in a _small_ (4qt?) sample over several hours running continuously. The cold side of the plate would frost up on its own, but cooling anything significant seemed to fizzle. I was hoping to circulate cooling fluid through a stainless loop that would be resting in the fermenter directly. Note that the air off of the radiator got plenty hot, too.

I initially tried using big fancy computer heat sinks (liquid filled heat pipes and shiny copper) with big jet engine sized fans (minor exaggeration - but they sounded like it) to radiate the heat away from the peltier - when that didn't work out well I was theorizing that the heat load of the peltier was keeping the liquid in the cpu cooler permanently in the gaseous state and reducing the overall cooling ability of the unit. That may be bs - but later reading lead me to believe that the cpu coolers are sized for the thermal load of the cpu, which I was exceeding by a good margin. I remember thinking that solid copper would be the way to go if not glycol.

The good heat sinks (not many of them) are rated in degrees above ambient temp per watt of energy disappated at a given CFM of airflow. Meaning you can directly calculate how hot the hot side will be when your peltier is working at its hardest. That should give you an idea of what temp the cold side will be, given that the TECs typically are rated for some particular hot-cold side temp difference. Not sure if that's available for your cpu cooling unit.

Also an issue for research - do computer power supplies give clean enough power for peltiers to perform efficiently? I seem to remember reading somewhere years ago (can't find it now) that the switching nature of pc power supplies was not ideal.

I suspect my issues were:
insulation - more more more
power - the leads out of my 600 watt psu got worryingly hot, although technically capable of putting out above and beyond the rated amps and voltage for the peltier
mounting - did I use the right amount of thermal compound and pressure when attaching the water blocks to the peltier?

I also looked at the Ice Probe for possible direct mounting into the side of a fermenter. Probably would need a few to get the job done. And for that matter, I've always wondered if the MoreBeer cooled conicals had their peltiers flush mounted on the outside or if there was some sort of probe extending inside...

Also looked at the Coolbot - a controller for using ac units for refreigeration. Good info on their site about DIY cooling chambers, although not targetted at fermentation specifically.

Apologies for rambling - I hope something in there is of interest

fish4fun

Recipes 

airving,

Thanks for the info! I am glad you have gone down this path before, and you mentioned something that is a concern:

While I have not found "good" specifications on the particular unit I selected as a starting place, I suspect it is designed to remove < 150W of heat (the most "power hungry" CPU in the supported sockets is rated @ 140W). As I am certain you know/discovered the "hot side" of the TEC has to dissipate both the "heat load" AND the heat generated through I^2R losses. The particular TEC I am starting with is rated @ 86W of cooling @ 136.5W input. This means the "hot side" has to dissipate 222.5W at full input power. The "critical temperature" for the type of TEC I am using is ~80C, and the maximum temperature difference between the two sides is ~65C.

Thermodynamics was not a class I did particularly well in back when I was young and sharp, and since all the math in the world simply gets you to a "safe region", my rather old and dull self is going to approach the problem from a purely empirical/experimental point-of-view. I do not expect the liquid cooler to be able to "keep up" with the TEC @ full power; however, The greater the temperature differential between the "hot side" and the ambient air, the more heat it will remove (technically: "the faster the heat will flow"). These facts have led me to one possible approach, the one I plan to explore first, which is "pulsing" the current to the TEC and using feedback to regulate the pulse magnitude and duration. (Really this is a bit mis-leading, technically what will actually happen is "pulses of pulses". That is, the current will be regulated by PWM (pulse width modulation) in the micro-second time domain, and the TEC's temperature will be regulated by periods of time when the PWM is On and periods when the PWM is Off, these periods will be in the second to minute time domain.) Following is a simplified version of the algorithm:

While I have NOT read anything to suggest that peltier coolers don't play well with "switching power supplies" in general, I can see how an unfiltered PWM pulse train might present a problem. I know that many people state that the 12V rail of a PC power supply will not fully drive a typical 15.2V TEC while a typical 12V car battery will (Nominal voltage of a typical lead-acid car battery is 14.4V despite its name plate rating of 12V). I do not expect this to be a problem, but if I do run into problems I will certainly look more carefully at the power supply.

Fish

airving

Recipes 

Check this link: Tellurex - Peltier FAQ

Items 40 and 41 are of interest with respect to voltage, discussing modulating a voltage above Vmax to bring it down to Vmax on average (#40), and modulating a good voltage to affect the duty cycle (#41).

My concern was case #1 of those two, but I suspect I'm off base here. What I've been able to find does not imply that the output of computer power supplies is anything but good solid dc power at the rated voltage (+/- up to 5%). I was worried that it was something like 30 volts sliced in such a way that it averaged to 12v or 5v, etc. Looks like no - so please disregard that issue.

fish4fun

Recipes 

airing,

What a GREAT reference! Thank You!

If initial testing goes as planned I do plan on building a line driven SMPS for the TEC. I am 99.9% certain that using a "fixed off time" current regulated buck converter with over-voltage sensing will be fairly easy to construct; it is very similar to high current LED drivers that I have built in the past (LEDs are a bit more voltage sensitive than TECs, and even in them maintaining a ripple current below 120mA is fairly easy to do).

FYI, PC power supplies are typically "Forward Converters" meaning they drive transformers. The voltage ripple in most decent PC power supplies is very good, much better than required by a TEC. They are only "cheap" because of the huge volume.

Fish

mattd2

Recipes 

I looked into peltiers a while ago (all in my mind though, never got to test anything ) and my biggest issue was getting the hotside cool enough that for a efficient peltier the delta_t was small enough to get a decent heat removal rate. Basically it was because you are relying on sinking your waste energy to ambient so to get enough heat transfer away from the hotside to ambient you need a large temperature difference. But if the hotside is hot enough to get decent heat removal from it it means the delta_T needs to be large to get a usable temperature on the cold side. Having a large delta_T means you have a much reduced heat transfer rate (power) through the TEC making it highly inefficient (I seem to recall one source claiming about 10% efficient, meaning that to get your 80W of cooling you need to load the thing up with 800W and then dissipate all that energy!)
Remembering either:
The max power is at delta_T = 0°C, i.e. it can only pull out the 80W if the hot and cold side are at the same temp (confusing concept )
or
The max delta_T is when the power= 0W, i.e. it will not remove any heat if the hot and cold sides are 65°C (still a confusing concept! )

Good luck!

Edit: a few HBT threads from the past:
http://www.homebrewtalk.com/f51/peltier-fermenter-cooler-using-water-heat-exchangers-138297/
http://www.homebrewtalk.com/f51/diy-thermoelectric-temperature-control-27524/

__________________
My DIY grain mill
First AG mini-BIAB

fish4fun

Recipes 

mattd2,

Great links and very good reading. Thank You! The positive results are close to what I am hoping for, the negative results are what my thinking/math suggested. My design calls for a highly insulated box (something the positive results seem to suggest is imperative). The negative results focused on directly coupling the controlled volume with the ambient environment (as opposed to liquid transfer). Despite the large number of negative results, I am very encouraged:-) I do not expect "magic", I expect a cooling system that is ~80% less efficient than a compressor based system, but at a fraction of the DIY construction cost. The threads you linked seem to suggest my expectations/calculations are realistic. Again, THANKS!

Fish

fish4fun

Recipes 

Everything always takes so much longer than I think it will, lol. Anyway, I have the uController code close enough that I felt it was time to assemble the TEC, Radiator/Pump/Fan and "cold sink", and "fire it up". I am only running the TEC @ 5V (roughly 17W input power). The "cold sink" is open to ambient. I assumed the "cold sink" would get cool to the touch and the air coming from the radiator/fan would be "warm". I was partially right. The "cold plate" quickly formed a layer of frost, but the air coming from the radiator/fan is indistinguishable from ambient. After running continuously for the past hour or so there is now a thick layer of ice covering the cold plate.

I am fully aware that forming ice on an uninsulated cold plate is not a good indicator of system performance, but it is encouraging none-the-less. After several hundred hours dedicated to code writing it is very nice to see some "real world signs" that so far the math and the performance seem to be in agreement. The wild-card was the "hydro-cooler" (radiator/pump/fan), as there were literally NO specifications available on it. The rather crude test I started today indicates it is far more efficient than I had thought it would be. I don't have a proper therm-o-well installed in my "cold sink", but sticking a thermometer on the surface indicates the "cold plate" is ~-6C (20F). Ambient air temp is 18.3C (66F), thus @ 5V // 17W I am maintaining a temperature differential between hot-side // cold-side of at least 25C (77F). The specs for the TEC state that it should be capable of dT of ~77C (170F) @ the rated power of 136W, but a great deal depends on the ability to remove heat from the "hot side", with the device critical temperature being ~85C (185F).

Some "odd notes" about heat transfer, in no particular order...The TEC's ability to achieve sub-freezing temperatures is not the determining factor in suitability for my proposed cooler. In the cooler itself there will be a second fan transferring heat from the inside of the cooler to the "cold plate". This will elevate the temperature of the "cold plate" considerably, but should also increase the efficiency of the overall heat transfer process. Increasing the power to the TEC may or may NOT lower the temperature of my "exposed to ambient cold plate". Initial indications are that PWM control of the TEC is effective (there were some reservations about that voiced earlier in the thread).

With another week or two (or if things keep going like they have been a Month, lol) of debugging left on the uController code, I am starting to think a simple PID Temp controller off ebay for $30 might really be all that is needed. The uController can simply control the TEC/Fans/Pump and leave the UI to the PID Temp controller (greatly simplifying the code. At this point the UI is PC based, adding buttons & an LCD to the uController would add months to development, yea, I know, I could simply go with an arduino....).

I am going to order one of the PID temp controllers off e-bay and turn my focus to "the box". I started a "box" built from 2" foam sheet from HD, but I think I am going to abandon it and build the mold for the pourable foam box instead. After-all, my goal is to build half a dozen of these, and I don't really want one to be an "ugly duckling" ;-)

Following are a few pictures of the TEC and my prototyping board....




This last one is a close-up of the TEC itself. Please note that the "white plate" the TEC is on is actually a brightly polished piece of aluminum 4" x 4" x 1/4" that is encrusted in ice! You can get some idea of how thick the ice is by the "scratches" I made in the ice using the "church key" in the other photos, lol.



Anyway, I am still working on this project, but there really had not been much to post about it up until now. I will post more as I make more progress.

Fish

fish4fun

Recipes 

I was so encouraged by the initial tests, I set the TEC up inside a 1cuft foam box with 1.75L of water (in a container). I put the "cold plate" and a fan inside the box and the radiator/fan on the outside. I sealed it up fairly tight, and fired it back up. The water started off @ 15.6C (60F). The ambient air temp is 19C (66F). Assuming I actually remove 17J/s of heat over the course of an hour (17Wh) (and assuming there is no heat gain from ambient, the fan or the pump), the water should drop ~8.4C (15F). Using the same formulas I used to calculate the heat gain of "the box" I plan to build, the heat gain through the 1in thick foam this 1ftx1ftx1ft box is made from should be ~7J/s when the temperature inside the box reaches 10C (50F). The minimum calculated temperature inside the box is -5C (23F) (with ambient @ 19C). This is the point where the heat gain = the heat removal rate (17J/s). In the half hour it has been running, the ambient air in the box has dropped to 12.2C (56F), but the water temp is still 14.5C (58F); this suggests an actual temp drop of ~4C in an hour with a temperature differential between the inside of the box and ambient of ~7.8C.

Well, I will know a lot more in a few hours....lol. While this is certainly NOT going to prove or disprove anything, it is much more indicative than watching Ice form on my "cold plate" ;-) I really want to push the TEC harder (75W to 100W), but I am going to hold off on that until I have better monitoring devices in place.

Fish

Fish

fish4fun

Recipes 

I got sloppy in my excitement and made lots of mistakes setting up my "test box". I compounded the problem by trying to continue by "patching mistakes" one at a time and then monitoring the temperature for a couple hours. At the end of almost two days of this approach I finally achieved 5C (41F) in my 1ftx1ftx1ft box, but it was a hollow victory, and I would have been much better served to stop "playing" and "do it right". To that end, I am going to back up a bit and properly design and build a "test box" and define the tests that will be truly indicative of system performance in a full scale design.

For a clear and concise read on the TEC math: http://www.directron.com/tecinfo.html

It will probably be late next week before I have much time to work on this again; I hate I wasted so much time on undefined tests, but the fact I achieved 5C inside the box is at least encouraging.

Fish

fish4fun

Recipes 

Ok, here we go :-)

The Test Box:
1ft x 1ft x 1ft box with 5 sides made from 1" foam and the lid made from 3/4" foam with TEC/Cold Plate/Fan/Liquid Pump mounted and sealed in the 3/4in foam. 1L of water in a plastic container inside the box.

The TEC/PS:
TEC1-12709 from China run @ a nominal 12V from a PC power supply.

I currently have only monitored the liquid Temp inside the box. The initial experiments were to determine the heat gain/loss of the box because until this is established nothing of interest can be learned about the TEC performance.

The following was used to predict heat gain inside the box:
An initial value of q was chosen from ( http://www.engineeringtoolbox.com/thermal-conductivity-d_429.html ) as 0.033.

Box Area = 0.54m^2
Thickness = 0.0254m

I used the TEC to pull the water temp down to 1.5C and then turned the TEC, Fans & pump off. I monitored the temperature of the water using a uController to log the temps over the course of the next 12 hours. The actual thermal transfer suggests:

q = 0.00948

This implies that the box has roughly 3.5 times LESS heat gain than predicted. I don't think I have "magic foam", so I suspect the air in the box is at some temperature between ambient and the water temp. What this means is that the transfer of heat from ambient to the air inside the box probably occurs at or near the initial calculated rate; however, the transfer of heat from the air inside the box to the water is delayed. Attempting to calculate the actual heat transfer function is a great deal more work than I am interested in doing, and quite honestly is pointless.

At the end of the day what we are actually interested in is liquid temperature and it's relationship to ambient. Knowing that the TEC is capable of removing enough heat from the box to lower the water temperature to 1.5C is all I really need from my 1ft x 1ft x 1ft box. I cannot think of a case where freezing temperatures are important, so all further tests will involve temps >2C.

The firmware for the uController is finally coming together very nicely. I ended up essentially writing an OS for the AVR that runs "scripts". These "scripts" can be run interactively via a PC interface or stored in EEPROM and be run in stand-alone mode. Of course this also required writing a PC interface, lol, but in the end being able to test and debug each function has saved my tail.

An interesting note about PC power supplies and the particular way I am using them is that the ATX standard calls for 5V standby power of at least 2A. What this means is that when the power supply is "off", there is still 5V available to run the uController. The uController can then "turn on" the ATX supply when a call for cooling or a call for heat is required. This means that when the TEC is off, the power consumption of the system is very, very low (less than $0.15 per month).

Initial power estimates for the TEC and the system are still all within initial design parameters.

Next step is a full Sanke keg size box. I know most are more interested in a carboy sized unit, and I will get to that in time, but since the only people interested want an "open source DIY project", I am going to finish my project first ;-) Summer is coming and I want to do some lagers :-) My brewing buddy/neighbor is a carboy brewer, and he is going to need to ferment in his garage now that his "spare room" has his new son living in it. In this hot part of the world he will need something to keep his carboys from cooking out there, so, I will do a full blown "how to DIY a TEC fermenter cooler for your carboy" some time this Spring. In the mean time I will report back when my "big box" is up and running!

Fish